Articulation and Finger Forces in Saxophone and Clarinet Playing (PDFDrive)

Articulation and Finger Forces in
Saxophone and Clarinet playing
Am Institut für Wiener Klangstil der

Universität für Musik und darstellende Kunst Wien
zur Erlangung des akademischen Grades eines
Ph.D. (Doctor of Philosophy)
im Fachgebiet der Musikalischen Akustik
eingereichte Dissertation
von
Alex Michael Hofmann
Datum der Einreichung (submitted): 29. 1. 2015

Datum der Disputation (defended): 11. 6. 2015
Datum der Veröentlichung (published): 17. 6. 2015
Begutachter:
Ao. Univ.-Prof. Dr. Wilfried Kausel
Univ.-Prof. Dr. Christoph Reuter
Abstract
Professional woodwind players practise for years until they are able to pro-
duce virtuosic performances. While performing, they have to control various
parameters at the same time, e.g., ngerings, blowing, lip position and tongue
articulation. This thesis puts a focus on tongue articulation measurements
and nger force measurements in saxophone and clarinet performance. Three
empirical studies as well as physical modelling sound synthesis address the
question of how players control the instrument to produce expressive tone
transitions.
First, in a sound production experiment on the saxophone and the clarinet,
the participants had to perform melodies which require dierent techniques
to play the tones (tongue-only actions, nger-only actions, combined tongue
and nger actions) in three tempo conditions. A strain gauge sensor-equipped
reed was used to investigate tongue actions of the players during performance.
Timing analysis of the captured data showed that in the slow tempo, a com-
bination of tongue and nger actions improved the timing. However, in the
fast tempo condition, the timing precision of combined tongue-nger actions
was close to the level of nger-only actions, which suggests that the ngering
1
technique has a dominant inuence on the timing of saxophone performances
and clarinet performances.
Second, nger forces applied to the tone holes of the clarinet were measured.
Clarinet students and professional clarinetists performed two tasks (expressive
performance task, technical exercise task) on a sensor equipped Viennese Clar-
inet. Although the individual nger force proles had overlapping tendencies
within the two groups of participants, the expressive performance task showed
higher nger forces. For the technical exercise task the mean nger forces were
lower. In particular, the group of professional players used the lowest nger
forces for the technical exercise task.
Third, in a listening experiment, it was questioned whether motor expertise
in music performance has an inuence on the ability to discriminate articu-
lation techniques in saxophone sounds. Listeners with dierent expertise in
1 published in Hofmann and Goebl (2014)

music making (saxophonists, musicians not playing the saxophone, and non-
musicians) participated in an AB-X listening test. Their task was to discrim-
inate saxophone phrases containing legato, portato and staccato articulation.
All participants could easily discriminate between staccato articulation and
portato articulation, whereas most errors occurred when the listeners tried to
discriminate between legato and portato phrases. In this case, the group of
saxophonists showed the best results, which indicates that expertise in saxo-
1
phone playing helped to facilitate the task .
Finally, the articulation parameters obtained during the experiments were
applied to a physical model in an attempt to simulate tongued and non-tongued
tone transitions.
1 published in Hofmann and Goebl (2014)

Zusammenfassung
Saxophonistinnen, Saxophonisten, Klarinettistinnen und Klarinettisten benöti-
gen jahrelanges Training um saubere Tonübergänge und üssige Läufe im
ausdrucksstarken Spiel erzeugen zu können. Während des Spielens haben sie
unter anderem die Finger, die Atmung, den Lippendruck und für bestimmte
Artikulationstechniken auch die Zungenbewegungen zu koordinieren. Saubere
Artikulation, ohne Nebengeräusche erfordert sehr viel Übung, da für einige
Artikulationstechniken die Zunge direkt das schwingende Blatt berührt. Der
Fokus dieser Arbeit liegt in der Untersuchung von Zungenartikulation und
Fingerkräften im Saxophon- und Klarinettenspiel.
In einer ersten empirischen Studie spielten Saxophonistinnen, Saxophon-
isten, Klarinettistinnen und Klarinetisten eine technische Übung, bei deren
Umsetzung verschiedene Spieltechniken koordiniert werden mussten (die Zun-
genbewegungen allein, die Fingerbewegungen allein, kombinierte Zungen- und
Fingerbewegungen). Es wurde ein spezielles Sensoreinfachrohrblatt entwickelt,
welches es erlaubt in den Blattschwingungen einen Zungen-Blattkontakt zu
erkennen. Die Ergebnisse konnten zeigen, dass in einem langsamen Spieltem-
po die Kopplung von Zungen- und Fingerbewegungen zu einer verbesserten
Rhythmik beitrugen. Allerdings war auch zu beobachten, dass im schnellen
Tempo die gekoppelten Bewegungen vorwiegend von den Fingerbewegungen
beeinusst waren. Somit kann man schlussfolgern, dass, obwohl die Zunge
den akustischen Tonbeginn steuert, die Finger einen stärkeren Einuss auf
1
die Spielpräzision haben .
Eine zweite Studie mit studierenden und professionellen Klarinettenspielerin-
nen und Klarinettenspielern fokussierte auf die Fingerkräfte beim Schlieÿen
der Tonlöcher. Eine an der Technnischen Universität eigens dafür entwickelte
Wiener Klarinette mit Kraftsensoren wurde für diese Messungen verwendet.
Obwohl die einzelnen Fingerkraftprole beider Gruppen groÿe Überlappun-
gen zeigten, wurde dennoch deutlich, dass ausdrucksstarkes Spielen zu einem
erhöhten Einsatz von Fingerkräften führte. Beim Spielen einer technischen
Übung wurden demnach im Mittel geringere Fingerkräfte gemessen und ger-
ade die Gruppe der professionellen Spielerinnen und Spielern zeigte in diesem
Fall den geringsten Fingerkrafteinsatz.
1 veröentlicht in Hofmann und Goebl (2014)

In einem Hörexperiment wurde untersucht, ob Expertise im Musizieren auch
zu einer verbesserten Unterscheidungsfähigkeit von verschieden artikulierten
Tonübergängen führt (legato, portato, staccato). Von den drei Teilnehmergrup-
pen (Laien, Musizierende die kein Saxophon spielen und Saxophonspielende)
zeigten die Saxophoninstinnen und Saxophonisten die besten Ergebnisse. Das
lässt die Schlussfolgerung zu, dass bei der Unterscheidung von Artikulation
in Saxophonmusik möglicherweise ähnliche Verarbeitungsvorgänge im Gehirn

1
stattnden, wie bei der Verarbeitung von Sprache .
Abschlieÿend wurden die sich verändernden, physikalischen Parameter, wel-
che eine Artikulation mit Zunge von der Artikulation ohne Zunge unterschei-
den auf ein physikalisches Model übertragen. Durch diese Resynthetisierung
der Klänge wurde verstärkt deutlich, welchen Einuss die Zungenartikulation
auf die Tonübergänge ausübt. Diese Einschwingvorgänge genauer zu studieren,
könnte ein Schwerpunkt für zukünftige Forschung darstellen.
1 veröentlicht in Hofmann und Goebl (2014)

Acknowledgements
This research was carried out within a funded FWF project on Measurement
and analysis of nger forces in clarinet playing (P23248-N24). The FWF
project was a collaboration between the Institute of Sensor and Actuator Sys-
tems at the Vienna University of Technology and the Institute of Music Acous-
tics (IWK) at the University of Music and Performing Arts Vienna. Parts of
this work were additionally funded through the MDW-Doktoratsstipendium.
I would like to thank, the leaders of this research project, Dr. Walter Smetana
and Dr. Werner Goebl, for giving me the opportunity to work within their
research groups.
First, I would like to especially give thanks to Dr. Werner Goebl for his im-
mense support and guidance throughout the last three years. Thank you for
discussing all my research ideas, the progress of my experiments, and for your
encouraging words during lunch and coee breaks. Furthermore, I am very
grateful to my supervisors Prof. Dr. Wilfried Kausel and Prof. Dr. Christoph
Reuter for supporting my research ideas and giving me helpful advise, espe-
cially with measurement techniques and signal processing methods. I am also
thankful to Dr. Vasileios Chatziioannou for the inspiring collaboration when
it came to physical modelling of articulation techniques and for the perfect
working atmosphere while we were sharing an oce. At this point I would like
to express my gratitude to Dr. Caroline Cohrdes for her support over all the
years and for her indispensable help on statistical methods.
The support I have received at the IWK was fantastic. In particular, I
would like to thank Alexander Mayer for his technical support in the labo-
ratory, Konstantin Zabranski for his support with signal processing, Laura
Bishop, Gerald Golka, Michael Weilguni, Saranya Balasubramanian, and Tat-
jana Statsenko for the inspiring discussions.
v
Furthermore, I could not have completed this thesis without the help of all
the participants who supported my research with their expertise, whether to
perform with the sensor instruments or listening to the sound stimuli. Espe-
cially, I would like to thank Prof. Oto Vrhovnik, and Dr. Barbara Schickbichler
for their support with organising the saxophone study and Dominika Knapp
for conducting the listening test. Additional thanks goes to Kerstin Höller and
Lila Scharang for their helpful advise when preparing the clarinet study and
for testing dierent sensor setups on the clarinet.
I am also very thankful to Katya Checkovich for proof reading parts of this
thesis and for her mental support in the last months of writing.
At this place I'd also like to say Thank you! to all my mentors over
the last decades without their passion for music, their knowledge about music
and their patience to teach music, I would not be able to write this thesis.
Jan von Klewitz, you ignited my passion for saxophone performance and your
encouragement made me nally decide to study music. Prof. Michael Beil,
thank you for opening my ears to new music, electro-acoustic compositions
and new media art. I am also very thankful to Prof. Dr. Stefan Weinzierl and
Prof. Douglas Repetto from Technische Universität Berlin for giving me the
opportunity to work as their teaching assistant.
Prof. Dr. Herbert Hellhund and Matthias Schubert at HMTM-Hannover,
thank you for giving me the chance to develop my personal approach to jazz,
live-electronics and saxophone performance and for giving me the freedom to
explore music and technology in parallel. Joachim Heintz from Incontri at
HMTM-Hannover, thank you for being my mentor in computer music, live-
electronics and sound synthesis, thank you for introducing me to the family of
Csound experts, and thank you for all your support and especially for being a
good friend!
Last but not least, I am profoundly grateful to my parents for supporting
me over all these years of education.

Contents
Abstract i
Zusammenfassung iii
Acknowledgements v
Contents vii
1 Introduction 1
1.1 State of the art . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Thesis outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Measuring musician-instrument interaction on single-reed wood-

wind instruments 5
2.1 Sound production on single-reed woodwind instruments and the
required player actions . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.1 Investigation of articulatory tongue movements in wood-
wind performance . . . . . . . . . . . . . . . . . . . . . . 6
2.1.2 Instrument shapes and ngerings . . . . . . . . . . . . . 7
2.1.3 Tongue and nger coordination in woodwind performance 9
2.2 Development of sensor reeds . . . . . . . . . . . . . . . . . . . . 10
2.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.2 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.3 Measuring articulation during performance . . . . . . . . . . . . 19
2.3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.3.2 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
vii
2.3.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.3.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.4 Finger force sensors on saxophone and clarinet . . . . . . . . . . 31
2.4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.4.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.4.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.5 General discussion . . . . . . . . . . . . . . . . . . . . . . . . . 40
3 Production and perception of legato, portato and staccato ar-

ticulation in saxophone playing 43
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.2 Experiment 1: Production task . . . . . . . . . . . . . . . . . . 46
3.2.1 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.2.2 Results and discussion . . . . . . . . . . . . . . . . . . . 52
3.3 Experiment 2: Listening test . . . . . . . . . . . . . . . . . . . . 60
3.3.1 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.3.2 Results and discussion . . . . . . . . . . . . . . . . . . . 62
3.4 General discussion . . . . . . . . . . . . . . . . . . . . . . . . . 63
4 Finger forces in clarinet playing 67

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.2 Performance experiment . . . . . . . . . . . . . . . . . . . . . . 70
4.2.1 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.2.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
5 Application of performance measurements to physics based

sound synthesis 93
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.2 Physical modelling . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.2.1 Modelling articulation on single-reed woodwind instru-
ments . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.2.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
5.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6 Conclusion and future work 101
6.1 Summary of contributions . . . . . . . . . . . . . . . . . . . . . 102
6.2 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
Bibliography 107
A Additional material saxophone experiments 119
B Additional material clarinet experiments 123
Chapter 1
Introduction
1.1 State of the art
Empirical research in music performance has been identied as a growing eld
in recent decades (Gabrielsson, 2003). There are multiple reasons that explain
this immense growth. On one hand, performance research provides detailed
insights into the characteristics of a musically professional performance and
helps to explain the interpretation process that happens when a professional
musician turns a written musical score into pleasant sound events. On the
other hand, performance research gives the opportunity to understand the
complexity of the underlying motor processes humans are able to perform in
order to control a musical instrument at a virtuosic level.
From the perspective of music acoustic research, the control that a player
has over a musical instrument is of particular interest, especially how cer-
tain player actions manipulate the resulting sound. The so called Musician-
acoustic instrument interaction has been named as a hot topic in recent acous-
tic conferences (e.g. in Stockholm Music Acoustics Conference in 2013).
Expressive performance on musical instruments requires coordinated and
goal directed body movements. These body movements have been categorised
according to their function in the process of performance (Leman and Godøy,
2010): Sound-producing gestures, communicative gestures, sound-facilitating
gestures and sound-accompanying gestures (Dahl et al., 2009). Video and
motion capturing technology is useful to observe and categorize player ac-
tions at instruments where the player-instrument interaction is visible e.g.,
1
2 Chapter 1. Introduction
with drum-set (Dahl, 2004), violin (Schoonderwaldt and Demoucron, 2009;
Rasamimanana, 2012), or piano (Goebl and Palmer, 2009, 2008). Jensenius
et al. (2010) dened a gesture space for piano performance. He dened player
actions to the keyboard as sound-producing gestured, footpedal actions as
sound-modifying gestured and dierent torso and head movements as ancil-
lary, sound-accompanying, or communicative gestures.
Clarinet performances have also been studied with motion capturing tech-
nology. Most of these studies focussed on the occurrence and manner of ancil-
lary gestures (Desmet et al., 2012; Caramiaux et al., 2012; Wanderley et al.,
2005; Palmer et al., 2009a). One study particularly focused on nger actions
in clarinet performance. Derived from a previous piano study in the same
laboratory (Goebl and Palmer, 2008), Palmer et al. (2009b) showed that clar-
inetists used larger nger movements and increased accelerations at nger-key
contact during dicult and fast musical tone sequences to achieve higher tem-
poral precision. There are still numerous of unanswered questions concerning
the ne motor control of ngerings in professional clarinet performance. (e.g.,
How much nger force do professional players apply to the tone holes? Is there
an optimal nger force prole?) Empirical data of professional clarinetists' n-
ger forces to the tone holes does not exist so far. In this research project an
empirical study to investigate nger force under dierent musical situations
is foreseen. Weilguni (2013) developed special ring-shaped force sensors for a
Viennese Clarinet at the Vienna Technical University.
However, most of the sound-producing and sound-modifying player-actions
in clarinet (and saxophone) performance happen at the mouthpiece. Such
embouchure actions include modications of the blowing pressure, the lip force,
and the lip position (Almeida et al., 2013), but also the control of the players'
vocal tract (Scavone et al., 2008), as well as articulatory tongue actions to the
single-reed (Liebman, 1989; Wehle, 2007).
A focus of this thesis is to measure articulation in clarinet and saxophone
performance. Which parameters describe the dierences between articulation
techniques (legato, portato, and staccato)? How does the player use the tongue
to modulate the reed oscillations directly? The aim of this research project
is to derive new tone onset parameters (and tone transition parameters) from
performance measurements. The obtained parameters will be used to extend

Chapter 1. Introduction 3
an existing physical model of a single-reed woodwind instrument (Chatziioan-
nou, 2010) towards the simulation of articulation techniques.
In woodwind performance, the player has to coordinate articulatory tongue
actions and nger actions to produce a sequence of tones. The question, re-
garding how far these dierent motor processes inuence the quality of the
performance is of interest for practising performers and music teachers. Simul-
taneous measurements of nger actions and tongue actions during performance
will provide new insights in sound-producing and sound-modifying gestures on
clarinet and saxophone. The results of the experiments are supposed to solve
unanswered questions about professional player-instrument interactions.
1.2 Thesis outline
This thesis is structured into four main Chapters (25). Each Chapter has its
own introduction where the relevant background literature is discussed. This
is followed by a report of my research and a discussion of the results.
Chapter 2 focusses on player-instrument interactions with single-reed wood-
wind instruments and discusses performance measurement methods. The pro-
posed measurement techniques put an emphasis on capturing the oscillations of
the single-reed, with nger actions at the tone holes simultaneously. Dierent
proposals to capture specic player-instrument interactions are presented and
evaluated. Finally, pilot measurements on the saxophone are presented, where
nger forces, key accelerations and reed signals are captured simultaneously.
1
Chapter 3 investigates the production and perception of dierent articula-
tion techniques on the saxophone with two empirical studies. In a production
experiment, 19 participants performed two melodies that required dierent
motor actions to perform the tones. Timing properties are investigated and
compared for the dierent performance conditions. In an adjacent perception
experiment, 31 participants with dierent expertise in music making had to dis-
criminate between audio clips with saxophone sounds of common articulation
techniques. The question of whether or not a background in music perfor-
mance has an eect on the perception of articulation techniques in saxophone
sounds is discussed.
1 published as Hofmann and Goebl (2014)

4 Chapter 1. Introduction
In Chapter 4, the sensor clarinet which was developed at the Vienna Uni-
versity of Technology (Weilguni, 2013), is used for an empirical study that
focusses on nger forces during music performance. In an expressive perfor-
mance task, excerpts from Clarinet Concerto No.1 in F minor (Op. 73) for
clarinet in Bb, by C.M.v.Weber were performed by 23 participants. In an ad-
jacent technical exercise task, the production experiment from the saxophone
study was repeated. Finger force proles, timing, and articulation techniques
of clarinet students and professional clarinetists are analysed in this data.
In Chapter 5, the results from the previous experiments (Chapter 2 and
3) are applied to a physical model of saxophone and clarinet. The aim is to
enhance the model towards the simulation of tongued and air-separated tone
transitions. Hereby, sensor reed signals from dierent performance situations
are used to derive meaningful parameter changes, which allow a modelling of
these two articulation techniques.
The research presented in this Ph.D. thesis was carried out within the
FWF project: Measurement and analysis of nger forces in clarinet playing
(P23248) at the University of Music and Performing Arts Vienna, under the
supervision of the national research partner Werner Goebl. To be in dialogue
with the international research community during the project time, prelimi-
nary studies were presented and discussed at eligible conferences ranging from
the International Symposium on Performance Science to the International
Congress on Acoustics. Reports, which include parts of this thesis, were pub-
lished in the associated proceedings and one study (Chapter 3) was published
as a peer reviewed research article in Frontiers in Psychology. The beginning
of each section in this thesis references published articles written by myself,
under the supervision of others. The footnotes provide explicit details about
my contributions in joint publications.

Chapter 2
Measuring musician-instrument
interaction on single-reed
woodwind instruments
In this chapter commonalities and dierences between the saxophone and the
clarinet are discussed with a focus on the resulting player-instrument inter-
actions. Methods to measure articulatory player actions and nger forces on
both instruments are evaluated to prepare the two empirical sound production
studies presented later in Chapters 3 and 4.
2.1 Sound production on single-reed woodwind
instruments and the required player actions
Clarinet mouthpieces and saxophone mouthpieces share similar construction
features. In both designs, a single reed of cane (or synthetic material) is
mounted to the bottom side of a beak-shaped mouthpiece (Nederveen, 1998;
Pinard et al., 2003). Subsequently, clarinet and saxophone mouthpieces are
controlled with a similar playing technique, where the player encloses the tip
of the mouthpiece with his lips, while the front teeth rest on the beak shaped
upper side of the mouthpiece. The lower lip covers the lower jaw teeth and
gives soft pressure to the reed. The controlling of teeth, lips and jaw is called
embouchure (Liebman, 1989).
5
Chapter 2. Measuring musician-instrument interaction on single-reed
6 woodwind instruments
During sound production the player's blowing into the mouthpiece, bends
the reed towards the mouthpiece lay. This makes the tip opening smaller, so
the amount of air entering the mouthpiece is reduced. At a certain point the
pressure that is built up in the instrument turns the reed motion towards the
other direction. Constant blowing from the player keeps the reed oscillating, at
a frequency which is related to the the impedance peak inside the instrument
body (Dalmont et al., 2003; Fletcher, 1979; Almeida et al., 2010).
In recent years embouchure related player actions were studied intensively.
Gazengel et al. (2007) showed that the lower lip has an important damping
eect on the reed's vibrations. They showed that without the damping of
the player's lip, the reed's own resonance is signicantly above the required
resonance for sound production on clarinet and saxophone. Consequently,
wrong adjustments of lower lip pressure from the player inuence the lip-
damping and may cause unpleasant squeaks in the sound.
Several research teams investigated the inuence of the vocal tract on the
sound of the saxophone (Scavone et al., 2008; Guillemain et al., 2010; Chen
et al., 2011). The researchers inserted pressure transducers into the the mouth-
piece and into the player's mouth, to measure the impedance spectrum inside
the mouth of saxophonists during performance. These measurements have
shown that in the high register, the vocal tract is crucial in adjusting ne
pitch corrections (Scavone et al., 2008). Moreover, Chen et al. (2011) observed
that controlling tones in the ageolet register (tones above the standard range
of the instrument) was only possible for advanced participants who were able
to adjust the resonance of their vocal tract to the targeted pitch.
2.1.1 Investigation of articulatory tongue movements in

woodwind performance
Tonguing is an important part of musical training and tongue articulation is
essential to produce expressive music on woodwind instruments (Bates, 1984;
Liebman, 1989; Koch, 1989; Pay, 1995). Clarinetists and saxophonists use
their tongue to directly modulate the oscillating reed to shape tone onsets and
tone osets.
For a long time, the examination of tongue motions during woodwind perfor-
woodwind instruments 7
mance has been an interest in the teaching of wind instruments. Medical equip-
ment, mostly from speech therapy, has already been used to monitor tongue
actions during performance. Gardner (2010) used ultrasonograhic imaging to
explain the function and movement trajectories of the tongue in professional
clarinet performance. He observed relatively stable tongue motion patterns
for recorded repetitions of two production tasks: speech production and clar-
inet articulation under dierent vowel and consonant conditions. With this
method, it was possible to show that in clarinet performance the tongue shape
remained stable across the entire pitch range of the instrument. A limitation
of this study was that all the data gained was from one recording session with
only one participant (the author himself ). Gardner also stated that even at a
slow playing speed, the frame rate of the ultrasound equipment (28 Hz) was
not ideal for capturing high speed tongue motions. Consequently, new meth-
ods to capture tongue actions during performance are required. A method to
capture tongue-reed interactions, using strain gauge sensors on the reed will
be discussed later in this chapter.
2.1.2 Instrument shapes and ngerings

In literature, the vibrating reed has been described as a mechanical oscillator
that acts as a pressure-controlled valve to modulate the ow into the instru-
ment body (Kergomard, 1995). The instrument body functions as a resonator.
Its impedance peak is determined by its length and shape. The reed oscillates
at a frequency related to this impedance peak. Tone holes (nger holes or key
holes) along the instrument body allow the player to modulate the frequency
of the tone. On the saxophone and the clarinet, ngers from both hands are
required to operate the tone holes. To play larger intervals, the player has to
close or open multiple tone holes at the same time. This requires simultaneous
nger movements, also called safe nger transitions. With a measurement on
the ute, Almeida et al. (2009) was able to show that unsafe nger transi-
tions in the dimension of a few milliseconds (21 ms) already inuenced the
radiated sound. Professional ute players were able to use smaller delay times
between their individual ngers compared to novice players. This had a posi-
tive inuence on their sound. Virtuous, expressive performance on woodwind

instruments requires nger movements with precise timing.
Professional musicians have to use their stamina eciently to be able to
play several hours per day and to avoid overexertion (Nemoto and Arino, 2007;
Spahn et al., 2011). Not much nger force is required to close a tone hole and
the force applied after the tone hole is closed has no inuence on the resulting
sound. Nevertheless, instrument teachers often reported that students use
too much nger force. Music education encourages loose ngering technique
(Wehle, 2007). Finger force measurements on clarinet instruments, foreseen
in this thesis, will gain new insights into the used nger forces of professional
players under dierent playing conditions.
The ngering schemata of clarinet and saxophone dier because of a main
constructional dierence between the instruments. Clarinets have a cylindrical
tube (instrument body), whereas saxophones have a conical tube. The reason
for this dierence comes from the fact that the size of the tone holes in the
instrument body has an inuence on the maximum dynamics of the radiated
sound (Hall, 1991). The predominantly cylindrical tube of the clarinet (cylin-
der with approximately 15.5 mm diameter) allows only small tone holes (312
mm). To build a single-reed instrument with larger tone holes, the shape of
the tube must be changed. Consequently, a conical tube of the same length,
as with the soprano saxophone (990 mm) allows larger tone holes (131 mm)
and can produce louder sounds than a clarinet (Nederveen, 1998).
Although both instruments (clarinet and soprano saxophone) share the
same excitation mechanism and their tubes are of approximately the same
length (650 mm), the shape of the tube dierentiates the two instruments. In
the sound of the saxophone, all members of the harmonic series appear, which
is a characteristic for the sound. Thus, overblowing to the rst harmonic al-
lows an octave shift of the tones from the lower register. On the saxophone,
the same ngerings are used to play in the higher register. Technically, the
large tone holes can not be covered by the ngers alone. The player has to use
the key pads which are located over the large tone holes.
Conversely, the sound of the cylindrical shaped clarinet includes only the
odd-numbered members of the harmonic series, which produces a dierent
timbre. Furthermore, this aects overblowing on the clarinet. Because of
the missing second harmonic, overblowing produces the sound of the third
harmonic, a fth beyond an octave. Hence, the clarinet requires at least 19
ngerings to produce all tones in the lower register and needs dierent n-
gerings for the overblown register. This has consequences on the clarinet key
work. Although, the small tone holes can be closed by the ngers directly,
additional tone holes are required for chromatic playing. These are coupled
to a complicated key work including rings around the tone holes and keypads.
Throughout the history of clarinet making (1770 until the early part of the
twentieth century), the clarinet key work was still improved (see Shackleton,
1995, for an overview). As a result, two slightly dierent ngering systems for
the clarinet are common today: the French (Boehm) system and the German
(Oehler) system.
In Austria, where this research project took place the Viennese clarinet
is predominantly played. Viennese clarinets share most of the construction
features of German clarinets, including the German (Oehler) ngering sys-
tem. The main dierence is a thicker instrument body with slightly larger tone
holes, in comparison to other German clarinets (Birsak, 2009).
2.1.3 Tongue and nger coordination in woodwind per-

formance
In saxophone and clarinet playing the ngers and the tongue have to be co-
ordinated. Depending on the musical situation, the ngers (legato-playing),
the tongue (tone-repetitions), or a combination of both (portato playing and
staccato playing) inuence the timing and the sound of the tones.
Recent performance studies on wind instruments focused mainly on one
single control parameter at a time e.g., the ngering technique (Palmer et al.,
2009b), the tonguing technique (Gardner, 2010), the lip pressure (Gazengel
et al., 2007), or the vocal tract inuence on the sound (Scavone et al., 2008).
Studies on tongue-nger coordination in woodwind performance do not ex-
ist. To gain information about tongue nger-coordination in single-reed wood-
wind performance, a measurement method that can capture tongue and nger
actions simultaneously is required. An important issue is that this technology
should not handicap the player, but capture reliable data and be suitable for
empirical studies. In this chapter, sensor technology applicable to saxophone

and clarinet are discussed, which allow investigations of tonguing technique
and ngering technique during performance.
2.2 Development of sensor reeds
This section describes a method to equip synthetic single reeds with sensors
in order to measure the oscillations of the reed during performance. Parts of

1
this section were already published in Hofmann et al. (2013a) . The follow-
ing section (2.2) is based on this article. The section also further includes
unpublished measurement results for clarinet reeds.
2.2.1 Introduction2
Saxophone reeds and clarinet reeds have a long tradition to be manufactured
from natural cane (Nederveen, 1998; Shackleton, 1995). The material prop-
erties of cane are well suited for this sound producing application, but the
disparity and sensibility of wooden reeds, with the resulting limits in operat-
ing time were often criticized (Liebman, 1989). The hygroscopicity of natural
cane leads to a variation in the elastic properties of the material. To be in
playing condition, traditional reeds have to be wet (Pinard et al., 2003). Dry-
ness worsens the sound quality. Chemists worked on synthesizing materials
which are applicable for a similar application but without these disadvantages.
In the following section the properties of synthetic reeds from the company
Légère are discussed, because these reeds were used in this research project.
Légère synthetic reeds are made out of an oriented polymer with material
properties (density: 0.91.1 g/mL; Young's modulus: 510 GPa) copied from
cane reeds in playing condition (Legere, 2000). The bending stiness of the
material determines the reed strength. The elastic modulus depends on the
1 Asthe rst author of this publication, I made the following main contributions: the
construction and the design of the sensor reeds, the design of the experiment, the data
analysis, and the writing of the research report. Calibration measurements were carried
out at the Institute of Sensor and Actuator Systems, Vienna University of Technology, in
collaboration with Michael Weilguni and Vasileios Chatziioannou. The procedure of the
experiment was under the supervision of Werner Goebl and Wilfried Kausel.
2 as published in Hofmann et al. (2013a)
molecular alignment of the polymer and can be regulated by the manufacturer
to produce reeds of dierent strength, thus all reeds are based on an identical
cut (like with cane reeds). Further eorts were undertaken to improve the
reed's sound properties. By keeping the polymer's density, while increasing
the Young's modulus (up to 16 GPa) plus making the reed 20% thinner, the
manufacturer reported that those reeds' sound was judged to be similar to the
sound produced by cane reeds (Legere et al., 2011).
Synthetic single-reeds gather several properties which favour them over nat-
ural cane reeds, especially for research applications. Non-hygroscopic reeds
provide better conditions for direct comparisons of measurements from dier-
ent time. Synthetic reeds are easy to clean and thereby suitable for empirical
studies. Besides, they have a longer lifetime, because the material is designed
to avoid splitting at the tip. Furthermore, lost or destroyed reeds can be re-
placed by replicates during an experiment because the material properties are
dened by the manufacturing method.
2.2.2 Method1
Strain gauge sensors
Bonded resistance strain gauges are based on the principle that the resistance of
an electrical conductor changes when it becomes stretched. Foil strain gauges
are produced by a low cost circuit printing technique and function as a sta-
ble, accurate, self temperature compensated measurement instrument. The
foil, with polyimide backing, allows a measurement lifetime of more than 105
cycles (Window, 1992). In this project, the sensor is used in a small opera-
◦
tion temperature range (2036 C), but in a wet environment. Hence, sensor
and wires have to be protected against moisture. The sensing area of a strain
gauge is a small (copper nickel alloy) grid on the foil. The sensor averages
the strain over this area by changing its resistance when bent. Consequently
a deformation of the sensor covered area of an object can be measured. To
capture the sensor's resistance several Wheatstone bridge circuit designs are
available for disposal (Window, 1992). In the current setup a quarter bridge
circuit, with 5V power supply and additional signal amplication (INA 126,

70 mm
65 mm
flat side
flat side
16.3 mm
12 mm
4 mm 6 mm
26 mm 30 mm
curved side
curved side
flat side sensor area

flat side sensor area 0.2 mm
0.2 mm 3.1 mm
3.1 mm
curved side
curved side
Figure 2.1: Schematic of an alto-saxophone reed (left) and a Bb clarinet reed, german
cut (right), equipped with a 2 mm strain gauge.
Figure 2.2: Synthetic alto-saxophone reed (left) and synthetic Bb clarinet reed, ger-
man cut (right), equipped with a 2 mm strain gauge.
by Texas Instruments Incorp.) was used.
Reed preparation
Standard industry foil strain gauge sensors (RS, 2 mm, 120 ohms) were at-
tached to the at side of a synthetic reed (see Figure 2.2). Attaching the
sensor to the side of the reed which shows inside the mouthpiece ensures that
during playing the player's lip does not touch the sensor. The sensor was
placed with some distance from the tip to avoid preparations at the thinnest
and most sensitive area of the reed.
The reeds were modied by the following four steps. First, the strain gauge
sensor was glued on the at side of the synthetic reed, with a distance of 4
mm (alto-saxophone reed) to the tip. Second, two 0.5 mm holes were drilled
near the heart of the reed, to lead the cables to the other side of the reed,
where a connection socket can be mounted. Third, the sensor and the cables
were protected with lacquer (transparent nail polish) against moisture inside
the mouthpiece chamber. Finally the drill holes were lled with hot-glue,
to make them airtight. The design properties had to be adjusted for the
Figure 2.3: Left: Clarinetist's embouchure while playing on a Bb clarinet. Right:

Clarinetist's embouchure on a Bb clarinet, sensor reed connector pins placed between
lips and ligature on the mouthpiece.
particular dimensions of Bb- clarinet and alto-saxophone reeds. See Figure 2.1
for respective dimensions.
Sensor calibration measurements

A universal testing machine (Inspekt micro LC 100N by Hegewald & Peschke,
Germany) was used to characterize the properties of a sensor equipped alto-
saxophone reed (Légère, strength: 2.25) and a sensor equipped Bb-clarinet reed
(Légère, German cut, strength: 4.25).
For the alto-saxophone, the sensor reed was mounted to a Vandoren (AL3,
1.52 mm tip opening) mouthpiece through the associated Vandoren Optimum
ligature. A customized mounting device held the mouthpiece tight to a motor
driven table of the testing machine (see Figure 2.4). Lifting of the table caused
the reed to press against the load cell (measurement head). The machine lifted
µm
the table by 20
s
, while a 100 newton load cell tracked forces occurring at its
tip (sampling rate: 10 Hz). During this process the strain gauge sensor signal
was captured simultaneously. For measurements with the clarinet sensor reed
a Maxton mouthpiece (NA-1, 0.75 mm tip opening) was used.
The sensor reed properties were measured under three dierent conditions.
In the rst condition the reed was bent by the resistance of the load cell,
µm
applied to the tip of the reed (20 for 1500 µm, Figure 2.5). In the second
s
measurement condition, a soft clamp was added to the setup to the position of
the player's lip (14 mm from the tip), to simulate an embouchure (Figure 2.6).
Figure 2.4: Clarinet mouthpiece with sensor reed, mounted to a universal testing
machine (Inspekt micro LC 100N by Hegewald & Peschke, Germany). Setup for
measurement condition one, bending the tip of the reed.
Again, the reed was bent by the resistance of the load cell, applied to the tip
of the reed. In the third measurement condition the load cell was applied to
the position of the player's lip, in an attempt to measure eects of the player's
lip to the sensor signal (Figure 2.7).
2.2.3 Results1
For all measurements, the displacement of the table, the measured voltages
from the strain gauge sensor circuit and the force at the load cell were captured
simultaneously, while the machine lifted the table.

Load cell
Displacement
Strain
gauge
Saxophone Reed Clarinet Reed

Displacement (mm) / Atrain gauge signal (V)
Displacement (mm) / Strain gauge signal (V)

1.5
1.5
Reed displacement (mm) Reed displacement (mm)
Strain gauge signal (V) Strain gauge signal (V)
1.0
1.0
0.5
0.5
0.0
0.0
0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0
Force (N) Force (N)
Figure 2.5: Measurement Condition 1: The load cell's (measurement head) resistance
bends the tip of the reed (top). Left: Force-displacement curve (solid, black line) and
strain gauge sensor signal (dashed, red line) for the alto-saxophone reed (strength:
2.25). Right: Force-displacement curve (solid, black line) and strain gauge sensor
signal (dashed, red line) for the clarinet reed (strength: 4.25).
Measurement at reed tip

In the rst measurement condition, the reed was bent through the resistance
of the load cell at the tip of the reed. Figure 2.5 shows the measurement setup
and the two graphs with the captured data for both reeds. The solid black line
depicts the displacement of the reed's tip in relation to the force at the load
cell, while the dashed red line depicts the measured voltage deviations of the
strain gauge sensor.
The diagrams in Figure 2.5 show that the measurements of the reed dis-
placement and the strain gauge signal overlap, until a certain point, where
the reed closes the mouthpiece completely (Force with approximately 1 N).
At this point a deformation of the reed was visually observed during the ex-
periment. The measurements show a linear reed displacement curve, until
the closing point. This linear trend was opposite from our expectations. We
expected a non linear force-displacement curve, representing the entire reed-
mouthpiece system, where the bending stiness is raised by the reed touching
the mouthpiece lay (beating, Dalmont et al., 2003). Instead the results of our
measurement show a force prole, similar to the spring constant of a free reed.
This can be explained by the shape of the load cell. In our measurements the
head acted as a pointing force to only the tip of the reed. As a result only the
tip was bend and the reed did not touch the lay. Accordingly, the measure-
ment shows the local spring constant of the tip of the reed. In a real playing
situation the player's embouchure bends the reed towards the mouthpiece lay
and the Bernoulli force from the oating air stream, acts on the whole reed
area. Under these conditions the reed would beat against the mouthpiece lay.
However, Figure 2.5 shows that the strain gauge sensor signal overlaps the
force-displacement curve, hence the strain gauge measures the bending of the
reed's tip. Similar observations were made for the clarinet reed (see Figure
2.5, right).
Measurement at reed tip with articial lip
In the second measurement condition, a soft clamp was added to the position
of the player's lip (14 mm from the tip). The lip clamping force was adjusted
to that the tip opening was reduced to 0.5 mm for the saxophone and 0.4
mm for the clarinet. The load cell was applied with 3 mm distance from the
tip of the reed. The diagrams in Figure 2.6 show that the measurements of
reed displacement and strain gauge signal show similar characteristics to the
rst measurements. Displacement and strain gauge signal behave linear until
the point where the reed closes the mouthpiece tip (Force with approximately
0.75 N). Although this measurement setup is closer to a real playing situation,
the pointing force to the tip of the reed does still inuence the measurement
results. The measured curve depicts the local spring constant of the tip of the
reed without beating the mouthpiece lay.

Artificial lip Load cell
Displacement
Strain
gauge


1.5
1.5
1.0
1.0
0.5
0.5
0.0
0.0
0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0
Force (N) Force (N)
Figure 2.6: Measurement Condition 2: Articial lip added to the setup. The load
cell (measurement head) bent the tip of the reed (top). Left: Force-displacement
curve (solid, black line) and strain gauge sensor signal (dashed, red line) for the alto-
saxophone reed. Right: Force-displacement curve (solid, black line) and strain gauge
sensor signal (dashed, red line) for the clarinet reed.
Measurement at the position of the lip

In the third measurement condition, the load cell was applied to the position
of the lip (14 mm from the reed tip), to measure the eect of the player's
lip force to the reed and particularly its inuence to the strain gauge sensor
signal (Figure 2.7, top). Under this measurement condition a non-linear force-
displacement curve was observed (Figure 2.7, bottom, solid line). Positioning
the load cell to approximately the middle of the free reed part, changed the
results of the measurements extremely. First, the displacement-force curve
shows a non-linear behaviour, which can be explained as the eective stiness
of the reed-mouthpiece system. Compared to the spring constant of a free

Load cell
Displacement
Strain
gauge


1.5
1.5
1.0
1.0
0.5
0.5
0.0
0.0
0 5 10 15 20 0 5 10 15 20
Force (N) Force (N)
Figure 2.7: Measurement Condition 3: The load cell (measurement head) bent the
reed at the position of the player's lip (top). Left: Force-displacement curve (solid,
black line) and strain gauge sensor signal (dashed, red line) for an alto-saxophone
reed. Right: Force-displacement curve (solid, black line) and strain gauge sensor
signal (dashed, red line) for a clarinet reed.
reed, the bending stiness is raised by the reed touching the mouthpiece lay
(beating Dalmont et al., 2003). Second, Figure 2.7 (bottom, dashed red line)
shows that while bending the reed at the position of the player's lip, the strain
gauge signal remained constant. This proofs that the sensor reed signal is not
aected by the lip force of the player's embouchure. However, when the reed
closed the mouthpiece tip, the deformation of the reed can be seen in the strain
gauge signal (Saxophone reed: force = 14 N, displacement = 0.6 mm; Clarinet
reed: force = 7 N, displacement = 0.4 mm).

2.2.4 Discussion1
This section explained a method to equip synthetic single-reeds with strain
gauge sensors in an attempt to measure reed bending during musical perfor-
mance. In a material characterization experiment with an universal testing
machine, reed displacement and strain gauge signal (reed bending) were mea-
sured, in relation to the applied force.
A force applied to the tip of the reed showed that the displacement curve
was correlated to the strain gauge signal. Applying the pointing force to the
position of the player's lip, revealed no changes in the strain gauge signal.
This observation suggests that the strain gauge signal gives useful information
about the bending of the tip of the reed. Further investigations may help to
relate the reed-bending parameter to the displacement of the reed's tip (tip
opening). Tip opening is an essential parameter when working with physical
models of single-reed woodwind instruments (Chatziioannou, 2010). Deriv-
ing reliable tip opening values from reed bending measurements would allow
direct comparisons of the reed behaviour in real playing situations with cal-
culations from physical models (Chapter 5). However, calibrating the sensors
on the reeds remains a complicated topic. Reasons for measurement errors
may be the position of the sensor on the reed and the size of the strain gauge
sensor in general. Conclusions about the vibrations of the whole reed and
further insights on the tip opening parameters are limited. Future work fore-
sees investigations with a high speed camera and an articial blowing machine
to gain additional information to further expose the question of reed sensor
calibration.
2.3 Measuring articulation during performance
This section will focus on the measured eects of articulatory tongue actions
to the vibrating single-reed, captured under real playing conditions. Parts of
this section have been presented and discussed at two international conferences
2
and were published in the associated proceedings (Hofmann et al., 2012a and
1 aspublished in Hofmann et al. (2013a)

2I am the rst author of this publication and made the following main contributions: the
design of the experiment, the data analysis, and the writing of the research article. The
Hofmann et al., 2013a). For this section of my thesis, previously unpublished
measurements with the clarinet were added.
2.3.1 Introduction 1
Woodwind players may choose from a large repertoire of articulation tech-
niques, that allows them to shape tone onsets and tone osets according to
the intended musical expression. One technique which is often used is portato
articulation. Portato articulation is produced by soft tongue strokes to the
reed, while the player's blowing is constant (Koch, 1989). For fast sequences
some professional players also use the technique of double tonguing. Instead
of a second tongue stroke to the reed, "the hump portion abruptly rises up
striking the roof of the oral cavity which in turn stops the air ow from the
larynx" (Liebman, 1989, p. 31). Liebman describes the resulting sound eect
as less eective on saxophone than on the ute.
Dierent attempts to explain the articulatory functions of the player's tongue
have been made in the past, in an attempt to develop more realistic models
for physics based sound synthesis. Two contrary explanations of the eects
of tongued articulation to the vibrating reed have been discussed lately. On
the one hand, Ducasse (2003) describes that a tongue stroke to the reed has a
damping eect to the vibrating reed and the force of the tongue changes the
equilibrium position of the reed. On the other hand, in the model of Sterling
et al. (2009) the tongue is understood as a gate in front of the mouthpiece tip,
which only prevents the air-stream to enter the mouthpiece.
This experiment investigates tongued and air-stream separated tone se-
quences, performed on alto-saxophone and Bb clarinet in an attempt to gain
detailed insights into the reed behaviour during such tone transitions.
measurements in the laboratory were carried out in collaboration with Vasileios Chatziioan-
nou. The procedure of the experiment was under the supervision of Werner Goebl, Wilfried
Kausel, and Walter Smetana.
2.3.2 Method
To examine the eects of articulation techniques to the sound excitation mech-
anism on saxophone and clarinet, three control parameters were captured dur-
ing performance: the blowing pressure from the player, the bending of the
reed and the resulting inner mouthpiece pressure. The study was undertaken
in two parts. At rst, measurements were done on the saxophone with a

1
simpler setup . For the second measurement the setup was enhanced. More
precise pressure transducers, similar to the ones in (Scavone et al., 2008) were
used, to study more details of the embouchure at the clarinet. As the focus of
this study are articulation techniques and the resulting reed behaviour, only a
single-microphone measurement (inside the mouthpiece) will be used. Other
research groups (van Walstijn and de Sanctis, 2014; Chatziioannou, 2010) used
multiple microphones to track the wave propagation inside the resonator.
Experimental setup
To capture the blowing pressure, the reed bending and the mouthpiece pres-
sure simultaneously during performance, the following setup was prepared for
2
measurements on the alto-saxophone : First, the blowing pressure of the player
was measured by a pressure transducer (Technoterm 5402), which was attached
to a probe that ranged into the player's mouth. Second, the reed bending was
measured with a synthetic, sensor equipped, saxophone reed as described in
Section 2.2. Third, the inner mouthpiece pressure was measured by a small
condenser microphone (40DP 26AS, by G.R.A.S.) inserted into the chamber
of the mouthpiece (AL3, by Vandoren) through a hole on one side. All three
channels were recorded onto computer hard disk (A/D conversion DAQ Lab-
View 2011, by National Instruments Corp.) using a sampling frequency of
11.025 kHz (16 bits). For this experiment the mouthpiece was only connected
to the alto-saxophone neck.
For measurements on the Bb-clarinet, an enhanced measurement setup was
choosen: Calibrated piezo-resitive pressure transducers were used to measure
precise pressure values (in Pascal) inside the mouthpiece and inside the players
1 Hofmann et al. (2012a, 2013a)

2 Hofmann et al. (2012a)
mouth. The recordings were made with a higher sampling frequency of 44.1
kHz (16 bit). A simple cylindrical tube was used as the resonator, in an at-
tempt to collect data applicable for direct comparisons of the measured signals
to the output calculated with physical model simulations (inverse modelling,
Chapter 5).
The experimental setup (see Figure 2.8) consisted of a clarinet mouthpiece
(Maxton NA-1) attached to a cylindrical tube (length l = 0.33 m; radius
r = 7.5 mm), two piezo-resitive pressure transducers (Endevco 8507C-2) and a

synthetic clarinet reed (Légère, German cut, reed strength 3.75) with a strain
gauge sensor attached to it (described in Section 2.2). Both piezo-resistive
pressure transducers were operated with the accompanying DC dierential
voltage amplier (Endevco Model 136). One pressure transducer was directly
inserted into the chamber of the mouthpiece through a drill-hole. The second
pressure transducer was connected to a probe, that ranged into the player's
mouth.
Figure 2.8: Clarinet mouthpiece with sensor-reed and two pressure transducers, cap-
turing blowing pressure and inner mouthpiece pressure.
Procedure
A professional woodwind player (author) performed two common techniques
of tone transitions, with mezzo piano dynamics. In the rst case (tongue-
separated tones) a portato sequence was performed, where the player used
the front area of the tongue to directly interact with the reed to separate the
tones (as described in Liebman (1989); recordings see Figure 2.9 Timespan:
3 − 5 s). In the second part of experiment (Timespan: 5 − 7 s) (air-separated
tones) the same tone sequence was produced, but this time the player fol-
lowed the instructions for the technique of double tonguing (raise the hump
portion of the tongue, to strike the roof of the oral cavity, which stops the
air ow from the larynx). Additionally a tone was played, where the blowing
was increased slowly, until the sound starts. The experiment was done with
the alto-saxophone setup rst and later repeated with the enhanced clarinet
measurement setup.
Figure 2.9: Recorded signals on the saxophone, showing the reed bending (top panel),
the normalized inner mouthpiece pressure (middel panel) and the normalized blowing
pressure (bottom panel) under two playing conditions: Tongue separated tones (35 s)
with a constant blowing pressure. Air-separated tones (57 s), with a varying blowing
pressure. The blowing pressure measurements verify the applied playing technique.
Figure 2.10: Normalized inner mouthpiece pressure (bottom) and reed bending (top)
for increased blowing until sound occurs on the saxophone.
2.3.3 Results1
Increasing blowing pressure
On the saxophone, a tone was played without using the tongue, by continuously
increasing the blowing until the sound starts. Then the player immediately
stopped blowing. The bending signal (Figure 2.10, top) shows that the blown
air bends the reed towards the mouthpiece lay. At a certain point (2.2 s) small
oscillations begin and increase, until a steady state condition of the tone is
reached. When the player stops to blow, the tone ends and the reed returns
to the rest position.
The same procedure was recorded with the clarinet setup which captures
the pressure values of blowing and resulting pressure in the mouthpiece. Figure
2.11 shows pressure values, measured inside the player's mouth (bottom panel)
and inside the mouthpiece (middle panel), in relation to the reed bending
signal. At the very beginning (Figure 2.11, 0.0 s 0.4 s), the player removes
the tongue from the reed and slowly increases the blowing pressure (02 kPa).
The increasing air ow bends the reed towards the mouthpiece lay (0.4s 1.5s).
At a certain point (1.5 s) the reed begins to oscillate. Reducing the blowing
1 as published in Hofmann et al. (2012a) and Hofmann et al. (2013a)

Figure 2.11: Measured reed bending signal (top panel), mouthpiece pressure (middle
panel) and blowing pressure (bottom panel). Player removes tongue from the reed
and then increases blowing into the clarinet mouthpiece until sound starts.
pressure leads to a release of the tone (3.0 s).
Observed signals during sound production
The sound excitation principle on single-reed woodwind instruments can be
observed with this measurement technique. Zooming into a steady state part
of the sound shows the interaction between reed motion and pressure inside
the mouthpiece. In Figure 2.12, reed bending and mouthpiece pressure are
plotted in one graph. When the player's blowing is bending the reed towards
the mouthpiece (begin of red curve), the reed reduces the tip opening. This
motion is additionally reinforced by the Bernoulli-force from the ow of air
into the mouthpiece. The smaller tip opening lowers the air pressure in the
mouthpiece (decreasing blue curve). At one certain point (e.g. 0.002 s), the
reections from inside the resonator (instrument) turn the reed motion to the
other direction and open the mouthpiece tip. Constant blowing from the player
makes the system oscillate.

1.0
Reed bending
0.5
Mouthpiece pressure
0.0
−0.5
−1.0
0.000 0.005 0.010 0.015 0.020 0.025 0.030

Time (s)
Figure 2.12: Normalized inner mouthpiece pressure and reed bending during a steady-
state tone on the clarinet.
Articulation
Blowing pressure, reed bending and inner mouthpiece pressure were captured
simultaneously for two contrasting articulation techniques, respectively tongue-
separated and air-separated tone transitions (Figure 2.13).
The blowing pressure signal shows that for tongue-separated tones, the blow-
ing from the player was approximately constant through the note transition
(0.2 s, Figure 2.13). In contrast, to produce air-separated tones the player
changed the blowing pressure during the note transition. The enhanced mea-
surement setup used for the clarinet gives absolute values of blowing pressure
at 4.5 kPa during tone production in mezzo piano playing. This value remained
constant for tongue-separated tone transitions but dropped down to 0.7 kPa
for an air-separated tone transition. The blowing pressure signal additionally
veries that the player followed the instructions for the two dierent playing
techniques.
Extracted parameters: Tongued tone transitions

When the player strokes the reed with the tongue (tongue-reed contact, TRC)
this abruptly bends the reed towards the mouthpiece. The soft tongue surface
dampens the oscillations of the reed (Figure 2.13, top left). This leads to a
sudden and short reduction of the pressure inside the mouthpiece (Figure 2.13
middle left). When the tongue releases the reed (TRR), the reed bounces back
a) b)
0.8
0.8
Reed bending
Reed bending
0.4
0.4
0.0
0.0
Mouthpiece pressure (kPa)

4
4
2
2
0
0
−4
−4
Blowing pressure (kPa)

5
5
4
4
3
3
2
2
1
1
0
0.0 0.1 0.2 0.3 0.4 0.4 0.5 0.6 0.7 0.8 0.9
Time (s) Time (s)
Figure 2.13: Measured reed bending signal (top panel), mouthpiece pressure (middle
panel) and blowing pressure (bottom panel) for a single note transition on the clarinet.
Tongue separated tones (left) and air-separated tones (right).
to the equilibrium position and continues to oscillate.
◦
Looking into the reed bending signal in detail, a 180 phase inversion at
TRC (Figure 2.15, 0.11 s) can be observed. A possible explanation of this
behaviour could be a change of the boundary conditions of the reed. When
the reed is oscillating it has only one non-free boundary condition at the point
were it is mounted by the ligature. The tip of the reed remains exible and can
move towards the mouthpiece. In the moment when the tongue strokes the
reed, the boundary condition changes (Figure 2.16). Now the reed is mounted
on both sides (ligature and tongue). When the returning pressure from inside
the instrument acts to the reed, it deforms it in a way that the pressure can
escape at the side of the reed. Consequently, while the tongue touches the
reed, an increasing of the mouthpiece pressure deforms the reed, which might
a) b)
−0.4
−0.4
Reed bending
Reed bending
−0.6
−0.6
−0.8
−0.8
−1.0
−1.0
1.0
1.0
Mouthpiece pressure
Mouthpiece pressure
0.5
0.5
0.0
0.0
−0.5
−0.5
−1.0
−1.0
0.0 0.1 0.2 0.3 0.4 0.0 0.1 0.2 0.3 0.4 0.5 0.6
Time (s) Time (s)
Figure 2.14: Measured reed bending for tongued tones (top left) and air-separated
tones (top right); Inner mouthpiece pressure for tongued tones (bottom left) and air-
separated tones (bottom right) on the saxophone.
explain the observed phase shift in the reed bending signal (Hofmann et al.,
2012a).
Extracted parameters: Air-separated tones

In the case when the player stops blowing into the mouthpiece, this loss of
the energy interrupts the oscillating system. In gure 2.13 (top right) on can
see a smooth fade-out of the reed bending signal. As a consequence, the inner
mouthpiece pressure shows a linear decrease (Figure 2.13 middle right). If the
player starts to blow again, the reed bends towards the mouthpiece lay and
starts to oscillate.
2.3.4 Discussion1
This section showed that dierent articulation techniques result in dierent
variation of the embouchure related parameters. It was shown, that strain-
1 as published in Hofmann et al. (2012a) and Hofmann et al. (2013a)

1.0
Reed bending
Mouthpiece pressure
0.5
0.0
−1.0 −0.5
0.0 0.1 0.2 0.3 0.4

Time (s)
Figure 2.15: Observed phase-shift at TRC; Inner mouthpiece pressure (blue) and reed
bending (red) high-pass ltered and normalized for a tongued tone transition on the
saxophone.
pb
e
n gu
To
ue
pb
ng
To
ue
ng
To pb
Figure 2.16: Blowing pressure closes the reed during sound production (top). Without
tonguing, the inner mouthpiece pressure bends the reed away from the mouthpiece
(middle). With tonguing the inner mouthpiece pressure deforms the reed (bottom).
pb represents the blowing pressure and p the pressure in the mouthpiece.
gauge sensors attached to woodwind single-reeds can capture the vibrations of
the reed during performance and give precise information about tongue-reed
interactions without environmental interference. The obtained reed bending
signal, mouthpiece pressure signal and blowing pressure signal are presented
in Figure 2.13 for the clarinet (Figure 2.14 for the saxophone), focusing on
a single note transition with alternatively use of tonguing or intermission of
the air-stream. These measurements conrm the tonguing model of Ducasse
(2003), where a tongue stroke to the vibrating reed has been described by its
eect to change the equilibrium position of the reed and its eect to damp the
oscillations.
Furthermore, a dierence in the emerging transients can be observed, de-
pending on the way that the player controls the oscillations of the instrument.
From the reed bending signal it was possible to deduce that tonguing results
in a stronger modulation of the reed vibrations, compared to the case of air-
separated tones. Dierences also appear in the envelope of the pressure sig-
nals. Soft tonguing resulted in a smoother waveform of the pressure signals
during the note transition. Observations for the eects of tonguing on the reed
behaviour were similar for saxophone and clarinet and the described charac-
teristics occurred for all measurements obtained during these experiments.
It should be noted here that the reed bending signal (depicted using the mea-
sured normalized voltages of the strain gauge sensor setup) does not directly
correspond to the displacement of the reed's tip (see Section 2.2.4). Neverthe-
less it provides useful information on the nature of the reed oscillations and
gives useful insights to articulatory player actions. The reduced complexity of
the clarinet setup (simple geometry of cylindrical tube) and the enhanced mea-
surement precision of the pressure transducers, allows a collection of physical
parameters, applicable for inverse modelling techniques described in Chapter 5.

2.4 Finger force sensors on saxophone and
clarinet
This section reports preliminary test measurements with sensor reeds and with
force sensors on the saxophone and the clarinet. It explains the preparation
of the measurement setup used for the two empirical studies presented later in
Chapter 3 and 4. Parts of this section were presented and discussed at two
international conferences and their reports were published in Hofmann et al.

1
(2012b, 2013c) .
2.4.1 Introduction
Besides embouchure related parameters, ngerings play an important role in
woodwind performance. Although the nger force to the keys does not inu-
ence the sound directly, insights about the ngering behaviour is of interest
for musicians, music teachers and also musician medicine (see Introduction
Chapter 4).
2.4.2 Methods
Test measurements with the saxophone2
In a rst attempt to measure nger forces to woodwind instrument keys, stan-
dard industry force sensors (Flexiforce, by Tekscan Inc.) were attached to the
pearls of the rst three saxophone keys with Petro Wax (by PCB Piezotron-
ics Inc.). This should allow to measure forces of left-hand index, middle and
ring nger during performance (Figure 2.17, left). The sensing area of such a
force sensor has a diameter of 9.53 mm, which covers the key-pearl and has a
thickness of 0.208 mm. Simultaneous test recordings of sound, reed-signal and
1I am the rst author of both publications and made the following main contributions:
the design of the experiments, the measurements in the laboratory, the analysis of the data,
and the writing of the reports. The laboratory setup was prepared in collaboration with
Alexander Mayer and Michael Weilguni. The procedure was supervised by Werner Goebl
and Walter Smetana.
2 as published in Hofmann et al. (2012b) and Hofmann et al. (2013c)
force sensors were done to measure tongue-nger coordination in saxophone
performance.
A test measurement showed that when the player pressed the keys at the
position of the key pearl, the sensors measured the applied force. Figure 2.18
depicts the signals of the recorded sound, the tongue action captured by the
sensor reed and the nger action captured by the force sensor. The force
curve (bottom) shows that the nger arrived at the key (FK) immediately
before the tongue contacted the reed (TRC). In the following time period,
while the tongue dampened the reed, the nger depressed the key until it hit
the instrument body and closed the tone hole (key-bottom, KB). Then the
player removed the tongue and this tongue action released the reed (TRR).
The required nger force to close that key was approximately 1 N, which
comes from the resistance of the key-spring. Afterwards a redundant force was
applied to the key while holding it closed (5 N).
An empirical investigation of nger forces on the saxophone was foreseen
with this method, but failed because of interrupted data (Hofmann et al.,
2012b). To further investigate the reasons of missing nger force data on the
saxophone, in a second test measurement, a small and light web-camera and
acceleration sensors (PCB 352C23, by Piezzotronics inc.) were mounted to
the saxophone (Hofmann et al., 2013c). From the video recordings, it became
obvious that the measured force values where inuenced by the position, the
ngers arrived on the key pearl. Some players closed the saxophone keys
with a technique, where the nger did not cover the key-pearl and thus also
not trigger the force sensor. Figure 2.19 shows nger positions of individual
players closing all left-hand saxophone keys to play the tone g' (Eb notation).
In the photos on the left side, one can see that these players covered the key
pearl with their ngertips and also triggered the attached sensor. But the
gures on the right side show that this was not always the case. Some players
only touched a part of the force sensor. From these observations, it becomes
obvious that the measured force values from players not touching the sensor,
are not comparable with these of players who covered the whole sensor area.
Furthermore it was dicult to extract timing information of the ngers from
such inconsistent force curves (Hofmann et al., 2012b).
The acceleration sensors which were attached to the moving parts of the
keys, tracked timing information of the ngerings on the saxophone, even if
the force sensor (Key Force 2) did not capture any data (see Figure 2.21).
Test measurement with the clarinet

As part of the FWF project Measurement and analysis of nger forces in clar-
inet playing (P23248-N24), special ring-shaped force sensors for a Viennese
clarinet were developed. Each sensor has three sensing elements, which allows
to measure applied forces from all directions. Figure 2.22 shows the clarinet
with dismounted keywork, so that the ring-shaped force sensors are visible.
Details about the design and the production of the nger force sensors can be
found in Weilguni et al. (2012, 2013) and in the related Ph.D. thesis Force
Sensors for the Measurement of Finger Forces in Clarinet Playing (Weilguni,
2013).
A test signal, captured for a tone transition with the sensor clarinet is
depicted in Figure 2.23. The signals from the three sensing elements of one
sensor ring were summed (bottom). The force curve shows that the nger
arrived at the force sensor while the tongue dampened the reed. The measured
nger force was approximately 0.5 N at the time when the nger closes the
tone hole. Then the player removed the tongue from the reed (TRR) and
applied more force to the tone hole, while holding it closed (2.5 N).
Figure 2.17: Left: Force sensors (by Flexiforce) on left-hand alto-saxophone keys;
Right: alto-saxophone with mouthpiece, sensor reed and clip microphone (C419, by
AKG Acoustics).
1.0
0.5
0.0
Sound
−1.0 −0.5
TRR
TRC TRC
−8000
TRR
−11000
Reed bending
−14000
5
4
3
2
KB
Finger Force (N)

1
●
(FK)
0
●
0.00 0.05 0.10 0.15 0.20 0.25 0.30
Time (s)
Time (s)
Figure 2.18: Recorded data for a portato tone transition on the saxophone: Audio signal (top panel), reed bending (middle panel) and
35
key force sensor signal (bottom panel). Tongue-reed (TR) landmarks are indicated by a cross in the reed bending signal. The nger force
curve shows the nger-key contact (FK), the arrival of the key to the instrument body (key- bottom with 1N ) and a redundant force
applied to the key while holding it closed (5 N).
36
Figure 2.19: Photos of ngerpositions of four saxophone students, playing g' on an alto-saxophone. Depending on their playing technique,
they covered the force sensors (left) or only touched a part of it (right).
Figure 2.20: Left: Player not covering the force sensor with left-hand index nger.
Right: Accelerometers attached to the key-work of the saxophone to track key move-
ments.
Figure 2.21: Sensor-reed signal (red), nger-force measurements (black) and addi-
tional tracked key-work acceleration (green), captured during performance. Accelera-
tion sensors provide precise timing information of the ngerings.
38
Figure 2.22: Viennese clarinet with dismounted key-work, equipped with 6 ring-shaped force sensors (Photo by Michael Weilguni).
1.0
0.5
0.0
Sound
−1.0 −0.5
2.3 2.4 2.5 2.6 2.7
1.0
TRR
0.5
TRC
0.0
Reed bending
−1.0 −0.5
2.3 2.4 2.5 2.6 2.7
2.0
1.0
Ring Sensor (N)

0.0
2.3 2.4 2.5 2.6 2.7
Time (s)
Figure 2.23: Recorded data for a tongued tone transition on the clarinet. Radiated sound (top panel), Sensor reed (middle panel) and
39
summed force for sensor ring 2, measuring the left hand middle nger force (bottom panel). The nger force curve shows that the nger
closes the tone hole while the player damps the reed with the tongue. Similar to the observations on the saxophone, a redundant force is
applied to the tone hole while holding it closed.
2.4.3 Discussion
First, a measurement setup to investigate ngerings on the saxophone was
presented in this section. It was shown that accelerometers at the key-work
of an alto-saxophone captured the timing of the key actions more precisely
than force sensors. Additionally it turned out that standard industry force
sensors on the key pearls were not a reliable method to capture nger forces
on the saxophone (Hofmann et al., 2012b). On the saxophone, there is no
direct inuence of the ngering position to the sounding outcome, because the
key-cushion closes the tone hole independent of the particular properties of the
nger action. The angle and the position of the nger-key contact was highly
variable for dierent saxophonists. This was conrmed by video capturing the
ngers during performance (Hofmann et al., 2013c).
Second, nger force signals from the sensor clarinet were inspected. On
the clarinet the player has to cover the complete tone hole with the ngertip
and close it properly. Otherwise it would inuence the sound. As a conse-
quence, the nger has to be placed on the ring-shaped force sensor developed
by Weilguni (2013). An application of this sensor clarinet measurement setup
will give new insights into the applied nger forces during expressive clarinet
performance.
2.5 General discussion
In this chapter, dierent approaches to measure interactions between the per-
formers and their instruments were discussed for the saxophone and the clar-
inet. At rst, a method to measure the reed oscillation during performance was
presented. Therefore, synthetic single-reeds were equipped with strain gauge
sensors. In an experiment to characterise the properties of the sensor reeds,
the reed displacement and the reed bending were measured. This helped to
determined that there is no inuence of the lip force to the sensor reed sig-
nal (Hofmann et al., 2013a). Subsequently, sensor reeds were used to observe
the eects of dierent articulation techniques to the vibrations of the reed
(Hofmann et al., 2012a).
Two dierent approaches, to capture nger actions of the player were dis-
cussed. First, the measurement of nger forces to the saxophone keys with
standard force sensors were examined. The recorded data was biased by in-
dividual nger-key positions of the players. To gain only timing information
from the ngerings, the usage of acceleration sensors was evaluated. For stud-
ies, where nger force measurements are not required, it turned out that ac-
celeration sensors can be a very reliable technique to investigate nger timing
for future research applications (Hofmann et al., 2013c). Second, nger force
measurements with the sensor clarinet (Weilguni, 2013) were investigated and
a setup design to measure nger forces and tongue articulation simultaneously
in clarinet performance was proposed.
A combination of the developed sensor reeds, together with the nger force
sensor clarinet turned out to be a suitable measurement tool for empirical
studies of woodwind performances.

Chapter 3
Production and perception of
legato, portato and staccato
articulation in saxophone playing
This chapter has been published as a research article (CC-BY) in Frontiers

1
in Psychology (Hofmann and Goebl, 2014) . It reports and discusses two em-
pirical studies that were carried out in the laboratory of the Institute of Music
Acoustics at the University of Music and Performing Arts Vienna. The rst
experiment examines dierent playing techniques on the alto-saxophone and
investigates the inuence of these playing techniques on the performed timing.
The second experiment is a listening experiment in which recordings of saxo-
phone sounds with dierent articulation techniques had to be discriminated.
1 This chapter is the full text of the journal article Hofmann and Goebl (2014), reviewed
by two independent international experts in the eld (Richard Ashley, Joe Wolfe). As the
rst author and the corresponding author of this publication, I made the following main
contributions: the design of the production experiment, the development of the measure-
ment methods, the technical setup of the laboratory, the recording of all participants, the
development of own data processing methods for sensor reed signals, the data analysis, the
script programming in R-statistics, and the writing of the research report. The listening
experiment was designed in collaboration with Dominika Knapp, who also conducted the
listening test and discussed the data in her unpublished diploma thesis: "Artikulation am
Saxophon", Jaurová (2013). Werner Goebl supervised the development of the research, gave
advise of statistical methods and provided guidance during the nal peer-review discussion
prior to the publication of the article.
43
Chapter 3. Production and perception of legato, portato and staccato
44 articulation in saxophone playing
3.1 Introduction
Producing expressive sound on single-reed woodwind instruments is a highly
sophisticated motor task, requiring coordination between the ngers, aural
cavity and respiration (Almeida et al., 2013; Chen et al., 2011; Scavone et al.,
2008). On the saxophone, a single reed of cane (or synthetic material), thinned
on one end, is attached to the bottom side of a beak-shaped mouthpiece (Ned-
erveen, 1998; Pinard et al., 2003). The player encloses the tip of the mouth-
piece with his lips and blows into the tip opening. During sound production,
the player's air stream excites the reed so that it oscillates related to the fre-
quency of the impedance peak inside the instrument body (Dalmont et al.,
2003; Fletcher, 1979; Almeida et al., 2010).
To perform expressively on woodwind instruments, the player may use a
range of parameters to shape longer sequences of tones such as onset timing
and tempo or the loudness of individual tones. An important dimension in
woodwind performance is tongue articulation, thus referring to the way the
tongue controls the shape of tone onsets, tone osets, and the connections
between tones (see Bengtsson and Gabrielsson, 1983; Liebman, 1989; Kraut-
gartner, 1982). Goolsby (1997) reported that more than 21% of professional
band rehearsal time is spend on instructions of articulation.
Articulation techniques on saxophone can be grouped in two main types:
tongued articulation techniques and articulation without tonguing. Legato ar-
ticulation does not involve tonguing and its sounding result is the smoothest
note transition. Herby, only changes of the ngerings determine the timing
and precision of note transitions. Professional ngering technique is required
to produce smooth and clean legato tone-transitions on wind instruments
(Almeida et al., 2009). For tongued note transitions, the intensity and du-
ration of the tongue stroke to the reed denes the sounding result. Portato
articulation is produced by soft tongue strokes to the vibrating reed, while
the player blows constantly. The sound of consecutive portato tones has been
described to be close to that of legato; the subdivision of tones is very sub-
tle. Liebman (1989) gives instructions to the technique of tonguing on the
saxophone as follows: it is the front portion of the tongue containing muscle
tissue which aps upward stroking the reed. The resulting eect is that the
articulation in saxophone playing 45
reed's motion and sound are momentarily stopped. The actual sounding of
the articulation comes with the release of the reed (p. 28). In contrast to
the soft sound of portato separated tones, staccato tones are sharp and short.
These are produced by placing the tongue immediately back on the reed after
the initial articulation. A consequence of these two dierent techniques of ar-
ticulation (tongued, non-tongued), the timing of the performance is controlled
either by tonguing or by the ngers.
Timing precision in the execution of complex movement patterns is essential
for musicians to produce rhythm in a sequences of tones. Palmer et al. (2009b)
investigated the inuence of nger trajectories on temporal accuracy in clarinet
performance, but restricted their focus on legato articulation. They reported
a positive relationship between peak accelerations of nger movements and
temporal accuracy of the performance, and concluded that tactile information
available to the ngers supports timing control, similar to observations made
with piano players (Goebl and Palmer, 2008). In woodwind performance, the
ngers have to be coordinated with tongue movements to produce expressive
sound. Studies based on isochronous tapping tasks showed that a coupling
of synchronous movements operated by multiple eectors improved temporal
stability. Experiments by Ivry et al. (2002) showed that synchronous tapping
with both hands improved temporal stability, compared to tapping with only
one hand. Additional foot tapping enabled further temporal improvements. In
the case of saxophone performance, when tongue and nger actions have to be
coordinated, the multiple eector advantage may also be the case. In line with
these ndings, we hypothesise that there is a positive inuence of combined
tongue-nger actions on the temporal stability in saxophone performance. In
this study we will investigate how dierent articulation techniques in single-
reed woodwind performance aect performance timing.
Perception and action in human motor control are strongly connected. The
motor theory of speech perception argues that human understanding of speech-
based auditory stimuli is based on the ability to recognize related vocal tract
movements required to produce equivalent sounds (Liberman and Mattingly,
1985; Galantucci et al., 2006). Neuroimaging studies have shown that brain
areas active in speech production are also active for speech listening (Fitch
et al., 1997; D'Ausilio et al., 2009; Fadiga et al., 2002). Similar observations
have been made for the production and perception of music (see Manto et al.
(2012) for an overview). Overlap in the neural regions active when professional
pianists listen to familiar pieces and the regions active when they perform these
pieces has been observed (Haueisen and Knösche, 2001). The link between
production and perception of music has been further discussed in the theory
of auditory-motor interaction for music making (Zatorre et al., 2007). Recent
research has shown that musicians are superior in judging asynchrony between
sound and body movements for performances on the instruments they master
than for performances on other instruments (Bishop and Goebl, 2014). Taking
into account that professional saxophone players practice over a decade to
acquire the skill level to produce fast tone sequences with uent articulation, we
hypothesise that this motor expertise may also improve the ability to perceive
articulation in saxophone performance.
In this paper, we investigate articulation techniques in saxophone perfor-
mances in two experiments. In a production experiment, we examine timing
measures in relation to eector combinations (tongue, nger, and both), n-
ger movement directions (pressing for tone onsets versus releasing for tone
onsets), and dierent articulation techniques (legato, portato, staccato). In
a second experiment, we test whether motor expertise in a particular eld
(i.e., saxophone performance) inuences the perception of dierent articula-
tion techniques in recorded saxophone sounds.
3.2 Experiment 1: Production task
3.2.1 Methods
Participants
Seven female and twelve male graduate saxophone students from the University
of Music and Performing Arts Vienna (N = 19, mean age = 23 years, range
1833 years) participated in this study. On average, the participants played
their instrument for 10.7 years (range = 4.520 years) and practised 1.9 hours
per day (SD = 0.97). Eleven saxophonists reported they play classical music
only, while the remaining eight participants perform usually as members of
jazz ensembles.
Experimental design
Two isochronous 24-tone melodies were designed for the experiment. Both
melodies consisted of the same elements (Figure 3.1): The rst part (note
number 18) is a tone repetition, produced by only tongue actions with no
change of ngerings. The following notes (924) require a sequential depression
(melody 1) or release (melody 2) of keys by left-hand ngers. Both melodies
were given as a score for alto-saxophone (sounding a major sixth lower than
notated), with additional portato, staccato and legato articulation instruc-
tions. In legato articulation tone repetitions are not possible to play, thus note
numbers 18 were omitted in the score.
Melody 1 Tongue only Fingers + Tongue
Melody 2
Figure 3.1: Stimuli used for the production experiment. Two 24-tone melodies in
E-at notation. Note numbers 18 require tonguing only. In melody one (top), note
numbers 924 require sequential key-depression by left-hand ngers. In melody two
(bottom) a sequential nger lifting is required to open the tone-holes of the instrument.
Equipment
The experimental setup consisted of a sensor-equipped alto-saxophone, a mi-
crophone, a digital metronome, and a multi-channel recording device. Strain
gauge sensors (2 mm, 120 Ohms) attached to synthetic saxophone reeds (by
Légère Reeds, Ltd.) were used to capture the bending of the reed during
performance (Figure 3.2). The strain gauge was part of a Wheatstone quarter
bridge circuit with 5 V (DC) power supply (Hofmann et al., 2013a). The sensor
reed, the microphone (C414, by AKG Acoustics) and the digital metronome
(KDM-1, by Korg Inc.) were connected via BNC cables to a multichannel
analog-digital converter (DAQ LabView 2011, by National Instruments Corp.)

70 mm
flat side
16.3 mm
4 mm
26 mm
curved side
flat side sensor area
3.1 mm 0.2 mm
curved side
Figure 3.2: Left: Synthetic alto-saxophone reed, equipped with strain gauge sensor
(2 mm), glued with 4 mm distance from the tip on the at side of the reed, to avoid
direct lip/tongue contact with the sensor. Right: Mouthpiece with sensor reed used
in the experiments to capture reed bending during performance.
to capture the signals simultaneously. All signals were recorded onto computer
hard disk (A/D conversion with sampling rate 11.025 kHz, 16 bit resolution).
Procedure
The experiment was conducted in accordance with the Declaration of Helsinki:
Participants gave written consent prior to the experiment, played under nor-
mal performance conditions, and received a nominal fee at the end of the
experiment.
In the beginning of the experiment, each player had to choose a synthetic
saxophone reed out of four dierent reed-strengths (Légère: 2.0, 2.25, 2.5,
2.75). All saxophonists were allowed to use their own mouthpiece but played
on the same alto-saxophone (77-SA, by Stagg). The metronome provided the
synchronisation signal on each quarter-note beat. The introduced tempi were
120 beats per minute (slow, IOI for eighth notes = 250 ms), 168 bpm (medium,
IOI = 178.6 ms) and 208 bpm (fast, IOI = 144.2 ms). All participants got a 5
minute warm-up, to practice the melody with the metronome at a slow tempo.
For the experiment, each participant played both melodies in legato, portato
and staccato articulation. They synchronised with the metronome for two
repetitions and continued playing when the metronome was muted, until the
melody had been played six times in total. We recorded two trials per tempo
condition, ordered from the slowest to the fastest. The experiment lasted for
approximately 1 hour per participant. In total 4644 tones were recorded per
player (2 melodies x 2 trials x 3 tempi x 3 articulations; containing 145 tones
for portato and staccato and 97 tones for legato). After their performances,
the participants lled in a questionnaire about their musical background and
the experiences with the sensor saxophone.
Data analysis
Sensor equipped saxophone reeds were used to capture the bending of the vi-
brating reed during human performance. Figure 3.3 shows typical sensor reed
signals in relation to the radiated sound under three dierent articulation tech-
niques. In legato articulation (Figure 3.3a) no tongue strokes were performed,
contrary to portato articulation (Figure 3.3b) and staccato articulation (Fig-
ure 3.3c). The tongue strokes to the vibrating reed are visible in the captured
signals, because the tongue presses the reed towards the mouthpiece lay and
thereby damps the reed vibrations. We dene two characteristic landmarks:
rst, a tongue-reed contact (TRC), when the tongue touches the reed, second,
a tongue-reed release (TRR), when the tongue releases the reed and initiates
the succeeding tone.
The data captured during the experiment contained more than 88,000 played
tones, which makes a manual transcription impossible. A multiresolution anal-
ysis (MRA) based on wavelet methods has been used successfully for the anal-
ysis of various time critical signals, ranging from medical data (i.e., ECG time
series, Percival and Walden (2006)) to transcriptions of drum patterns in audio
recordings (Tzanetakis et al., 2001; Kronland-Martinet et al., 1987; Paradzinets
et al., 2006). The following section discusses a landmark detection function
(LDF) based on a wavelet decomposition of the sensor reed signal, where the
external libraries (wmtsa, msProcess) were used in the R-statistics software
package (R Core Team, 2012).
The reed signal was decomposed using the Maximal Overlap Discrete Wavelet
Transform (MODWT) of level J 0 = 11. A Daubechies least asymmetric 8-tap
lter LA(8) allows direct reference from the MODWT details to actual times
in the reed signal. Figure 3.4 shows the algorithm of the LDF, working in two
main steps. First, extrema in detail e 11 (time resolution: 4t = 92.88 ms) were
D
labelled. These extrema represent reed displacements caused by the tongue.
Hereby, maxima of D
e 11 were labelled as TRR, because the following signal
decrease is an indicator that the player released the tongue. Minima were
labelled as TRC because a contact with the reed must have happened before
releasing the reed. Second, landmarks were shifted to the extrema in details
with a higher time resolution (D

e 10 , D
e9 and e 8 : τ8 4t = 11.61 ms).
D A spe-
cial treatment of the legato recordings was required to locate tone transitions
without tongue actions. To ensure comparable detection results, the same
MODWT analysis was applied to the legato recordings, but with an adapted
LDF which worked on details D

e 10 , D
e9 and D
e8 only. One participant's data
had to be omitted completely from the analysis, as the sensor data indicated
that no tonguing was used in any of the playing conditions.
To evaluate the quality of the LDF, it was tested on a small data set which
contained 2020 manually annotated landmarks. Starting from the annotated
ground truth, the existence and number of detected landmarks around the
annotated events was checked. A single detection was counted as one true
positive, whereas double detections were considered as one true positive and
one false positive. A missing landmark was counted as one false negative. Re-
maining landmarks, not matched to annotated events, were counted as false
positives. The standard measures precision, recall and F-measure were used.
Recall describes the completeness of the search and precision gives status about
the quality of the search results. F-measure combines the two previous mea-
sures.
precision = true positives/(true positives + false positives)
recall = true positives/total number of items (3.1)

precision · recall
F − measure = 2 ·
precision + recall
Overall, the wavelet-based analysis gave satisfactory results of the detection
tasks with F-measure > 94% (see A.1 for detailed values, and Hofmann et al.
2013b for further discussions of landmark detection in the sensor reed signals).
To check possible inuences of the LDF to the regularity of the extracted
landmarks, we calculated the time dierences of all detected landmarks to the
manually annotated landmarks of the ground truth data set. A mean deviation
of 0.42 ms (SD = 6.84) showed that the detected landmarks were close to the
annotated landmarks.
Score-performance matching1
However, from the detected landmarks, the correct position of each perfor-
mance event in the score has to be determined. Score-performance matching
has been identied as a complex problem, especially if there are dierences
between score and performance data (Gingras and McAdams, 2011; Grachten
et al., 2013). Possible reasons for these dierences can be drastic changes of
expressive timing or mistakes that happened during the performance (omit-
ted tones, wrong tones). Additionally, in our case there may be errors from
the landmark detection function (false positives, true negatives) in the per-
formance data. Thus, it is important to verify each detected landmark, by
matching it successfully to an event in the score and to omit false detections
from further analysis.
A main problem in the score-performance matching with the saxophone
data was, that some players omitted tones to take breath within the phrases.
Additionally the F-measure of 94 % from the LDF constitutes that there may
be (about 89) false detections in each trail with 145 tones. To identify the
errors and exclude them from further analysis, a pattern-matching algorithm,
based on dynamic time warping was used to match the performance data to
the score. The technique of dynamic time warping allows to map events of two
sequences, even though if both sequences are of dierent length.
Appendix A.1 and A.2 shows two examples of the used procedure to map
the performance events to the score: First, an IOI similarity matrix between
performance timing and score timing was created, based on the time dierences
between subsequent events (inter-onset intervals: IOI, in ms). Second, pitch
information from the sound recordings of the performances were extracted with
SonicVisualizer software using the Brossier (2006) pitch detection algorithm.
With these pitch information, a pitch similarity matrix was created based on
the pitch class distances of subsequent events from both, performance data
and score information. In a third step, both matrices were combined, by mul-
tiplication. To create a nal cost matrix, the combined IOI × pitch similarity
1 This
subsection has been added for this thesis and is not part of the original article
(Hofmann and Goebl, 2014).
matrix was subtracted from an all-one matrix.
In a nal step, the dynamic time warping algorithm nds a path trough
the nal cost matrix which has the lowest costs. A step pattern determines
the rules that are allowed to the matching model. In this case, a Rabiner and
Juang (1993) step pattern was chosen, which allows to omit up to three events
of the sequence. With this procedure all detected events in the performances
were veried through mapping these to events in the score.
3.2.2 Results and discussion

Timing of performed melodies
To examine timing of the produced sequences, we calculated inter-onset inter-
vals (IOI, in ms), as the time interval between two subsequent TRR (onset)
landmarks (IOI x = tx+1 − tx ). From these IOIs we calculated the timing
error (accuracy) and the coecient of variation (CV, precision). The timing
error (IOI obs − IOI exp )/IOI exp describes the relative deviation from the given
tempo. A negative value corresponds to a sequence played too fast; a positive
value to a sequence played too slow. The temporal precision of the played
melody was calculated (CV = SDIOI /M eanIOI ) to examine the regularity of
the tone-events. CV values close to zero correspond to high regularity in the
sequence, while larger values indicate higher variability in the onset distribu-
tion.
The average signed timing error of all performances during the synchro-
nisation phase was close to zero (M = 0.0077, SD = 0.032). Figure 3.5a
(solid line) shows that all participants were able to play the melodies to-
gether with the metronome click in all three tempi. A two-way repeated
measures analysis of variance (ANOVA) on timing error by tempo condi-
tion (metronome IOI = 250 ms, 178.6 ms, 144.2 ms) and synchronisation
condition (with metronome = synchronisation, without metronome = contin-
uation) indicated a signicant main eect of tempo [F (2, 34) = 13.23, p <
2
0.001, η = 0.662] as well as a signicant interaction between tempo and syn-
chronisation [F (2, 34) = 35.77, p < 0.001, η 2 = 0.823]. Without metronome
click, participants increased the playing speed in the slow tempo condition
and reduced fast tempi to a more comfortable playing speed (Figure 3.5a,
Figure 3.3: Alto-saxophone sensor reed signals and radiated sound recorded in an
anechoic chamber, showing a note transition (d2e2d2) under a tempo instruction of
250 ms inter-onset interval (audio sampling rate 44.1 kHz). Examples are taken from
the pool of stimuli for the perception experiment in Section 3.3. a) Reed signal for
legato articulation without tonguing (red) and radiated sound (black); b) Reed signal
(blue) for portato articulation with tongued note onsets (tongue reed release, TRR)
and note osets (tongue reed contact, TRC) and radiated sound (black); c) Reed
signal (green) for staccato articulation, with extended tongue reed contact duration
and radiated sound (black).
−3000
Reed signal
−7000
200
D11
−400
−400 200
D10
200
D9
−300
D8
0
−400
−200 100
D7
0
D6
−400
−1500 1000
D5
0.0 0.1 0.2 0.3
Figure 3.4: Maximal Overlap Discret Wavelet Transform of a sensor reed signal
containing tongued articulation: The gure shows the input signal (top) includ-
ing detected landmarks (TRC: red circle, TRR: green circle) and the details of the
wavelet decompositon D
e 11−5 (below). The landmark detection function labelled max-
ima (green) and minima (red) in detail D
e 11 . These positions were rened to extrema
of D
e 10 , D
e 9 and D
e 8.
dashed line). Similar observations have been reported for performances on
other instruments (e.g., piano performance, Goebl and Palmer 2013). The
overall temporal precision of the played sequences CV = (SDIOI /M eanIOI )

was high (mean CV = 0.11). The same two-way ANOVA was calculated for
the CV and revealed a signicant main eect of synchronisation condition
[F (1, 17) = 19.61, p < 0.001, η 2 = 0.732] and a signicant interaction between
tempo and synchronisation [F (2, 34) = 6.92, p < 0.01, η 2 = 0.538]. Figure
3.5b (dashed line) shows the reduction of timing precision even for moderate
playing speeds when the metronome click was removed.
Timing with multiple eectors

The melodies (Figure 3.1) were designed to consist of three distinct parts which
had to be played with the ngers only (legato), the tongue only (portato note
a) Timing accuracy b) Timing precision
0.10
0.20
(too slow) ● Continuation ● Continuation
Mean signed timing error ● Synchronisation ● Synchronisation
0.05
0.15
Mean CV of timing
●
●
● ●
●
●
0.00
0.10
● ●
● ●
●
−0.05
0.05
−0.10
0.00
(too fast)
Slow Medium Fast Slow Medium Fast
Tempo condition Tempo condition
Figure 3.5: Timing error (a) and coecient of variation (b), for synchronization-
continuation playing conditions. When playing with metronome click (synchroni-
sation phase, solid line) and without metronome click (continuation phase, dashed
line). Error bars show the standard error of the mean.
repetition), and with tongue and ngers in a coordinated fashion (descending
and ascending note sequence in portato articulation). For this analysis we
restricted our data set to legato and portato recordings, because the onset
detection for staccato melodies was less robust.
We grouped parts of the melodies according to the eectors required for
playing and compared onset timing between these parts. A two-way repeated
measures ANOVA on timing error by eector combination and tempo, indi-
cated a signicant main eect of the executing eector [F (2, 34) = 25.05, p <
0.001, η 2 = 0.772], a main eect of tempo [F (2, 34) = 22.76, p < 0.001, η 2 =
0.757], as well a signicant interaction between eector and tempo [F (4, 68) =
28.39, p < 0.001, η 2 = 0.791; Figure 3.6a]. A post-hoc pairwise t-test veried a
signicant inuence of each eector condition on the timing error (Bonferoni:
p < 0.001). Playing with only nger actions led to faster performances than
the metronome in all three tempo conditions (mean timing error = −0.017).
Tongue-only actions led to slower performances compared to the metronome
(mean timing error = 0.026), especially in the medium and the fast play-
ing conditions (M = 0.045). Using both eectors in a coordinated fashion

(tongue + ngers) stabilised the timing error (M = 0.013), but showed a
signicant tempo reduction in the fast tempo condition (M = 0.058). We ob-
served a learning eect in the recording of the second trial for the same task
[F (1, 17) = 6.55, p < 0.05, η 2 = 0.527; see Table 3.1]. The timing error for
combined tongue-nger actions at medium and fast tempi was signicantly
reduced in the second trial. This indicates that professional players already
improved their tongue-nger coordination after the rst six repetitions and
were able to perform the second trial with reduced timing error.
A two-way repeated measures ANOVA on temporal precision (CV) by ef-
fector combination and tempo, showed a signicant eect of tempo [F (2, 34) =
2
5.76, p < 0.01, η = 0.503], and an interaction between tempo and the used
eectors [F (4, 68) = 6.57, p < 0.001, η 2 = 0.528; Figure 3.6b]. Looking at the
CV values plotted in Figure 3.6b (dotted line), we see that timing precision for
the nger-only condition was lower in the slow tempo condition than in the fast
tempo condition. A similar pattern appeared for the tongue-nger condition.
Contrary, tones played only by tonguing (solid line) showed almost a constant
irregularity over all three tempo conditions. This was conrmed by three sepa-
rate one-way repeated measures ANOVAs on the CV by tempo condition. The
results showed a signicant main eect of tempo for both conditions where
ngers were involved [ngers only: F (2, 34) = 13.25, p < 0.001, η 2 = 0.662;
tongue + ngers: F (2, 34) = 5.51, p < 0.001, η 2 = 0.495), but no signicant
eect of tempo under the condition of playing with the tongue alone. Sep-
arate post-hoc pairwise t-tests, three for each tempo condition, showed that
at the slowest and fastest tempo the tongue-only condition was signicantly
dierent from the other two conditions (Bonferoni: p < 0.05), except from
combined tongue-nger actions in fast tempo (p = 0.07). There was no signif-
icant dierence between the nger-only condition and the combined tongue-
nger condition across all tempi. These ndings suggest that at slow tempi
timing improves with combined tongue-nger actions. The observed eect can
be explained by the multiple-timer model (Ivry et al., 2002), where the timer
responsible for the tongue movements is coupled to the timer of the ngers.
Such a coupling of multiple eectors has been shown to improve timing pre-
cision. We found the opposite eect for the fast tempo condition: Tongue
timing deteriorated, when combined with nger movements. This fast tempo
0.10
0.20
(too slow) ● Fingers only ● Fingers only
● Fingers + tongue ● Fingers + tongue
Tongue only Tongue only
Mean signed timing error
●
0.05
0.15
Mean CV of timing
●
●
●
0.00
0.10
●
● ●
● ●
●
●
−0.05
0.05
−0.10
0.00
(too fast)
Figure 3.6: Timing error (a) and coecient of variation (b), grouped by eectors
used to produce tone onsets. Error bars show the standard error of the mean.
condition (IOI = 144 ms) examined performances close to the synchronization
threshold of professional musicians (100120 ms Repp, 2005). The measured
CV values were about the same level as tones played with only ngerings. It
is interesting that professional saxophonists were able to produce coordinated
movements under this extreme tempo condition, but did not benet from the
coupling of the tongue to the ngers. Hence, saxophonists' tongue movements
were coupled to the nger movements, even if the precision of the nger move-
ments was worse than the precision of the tongue alone. This indicates that in
saxophone playing, the timing precision of the ngers dominates the precision
of the overall performance, thus overruling the timing eects of the tongue.
Timing and the direction of nger motion

To play descending tone sequences on the saxophone, keys have to be pressed,
while ascending sequences require ngers to open tone holes. To see if the
direction of nger movements (pressing down versus lifting up) inuences the
timing of the performance, we contrasted (legato) sequences with a focus on key
depression (Melody 1) to those focussed on lifting the keys (Melody 2). A two-
way repeated measures ANOVA on timing error showed no signicant eect

Tempo (IOI) Slow (245 ms) Medium (178.6 ms) Fast (144.2 ms)
Trial 1 Trial 2 Trial 1 Trial 2 Trial 1 Trial 2
Fingers Only -0.03332 -0.03430 -0.00095 -0.01010 -0.01170 -0.01402
Fingers w. Tongue -0.02675 -0.02760 0.01610 0.00361 0.06659 0.05029

Tongue Only -0.00666 -0.00521 0.04549 0.03438 0.04636 0.04446
Table 3.1: Timing error for both trials of sequences performed with dierent eectors
in three tempo conditions. Bold numbers depict improved timing error (learning
eect) for combined tongue-nger actions at medium and fast playing speeds for the
second trial.
of the direction of nger motion nor any interactions with tempo. The same
ANOVA on timing precision showed no signicant eects. Similar observations
have been reported for clarinet performances by Palmer et al. (2009b).
Characteristics of articulation techniques

We recorded reed signals of melodies with legato, portato and staccato ar-
ticulation (Figure 3.3). Whereas for legato articulation, no tongue actions
were required, portato and staccato note transitions required precise tongu-
ing. Each of these articulation techniques allows variation within itself based
on the onset and oset timing. Bengtsson and Gabrielsson (1983) discussed the
resultant concepts of duration and emphasized the importance of controlling
onset and oset parameters for the motion character of the rhythm. We in-
vestigated the tongue-reed contact duration (TRdur) for portato and staccato
tone transitions by subtracting the TRC times (oset of previous tone) from
TRR times (onset of subsequent tone, Figure 3.3). The average contact du-
ration for portato articulation for all participants was 25.5 ms (SD = 4.1 ms,
see Figure 3.7). A one-way repeated measures ANOVA on TRdur by tempo
condition showed no signicant eect of tempo (F (2, 34) = 1.3, p = 0.295).

On the contrary, the same ANOVA on staccato articulation showed a highly
signicant inuence of tempo condition on the tongue-reed contact duration
(F (2, 34) = 25.2, p < 0.001). In staccato articulation, the contact duration
(gap between tones) varies with the tempo. We calculated the relative gap
duration for each note transition by T Rdur/IOI exp and calculated a one-way
0.10
Tongue−reed contact duration (s) ● Portato articulation
● Staccato articulation
0.08
●
0.06
●
0.04
● ● ●
0.02
0.00
Slow Medium Fast
Tempo condition
Figure 3.7: Tongue-reed contact duration under dierent tempo conditions for portato
articulation and staccato articulation. Error bars show the standard error of the
mean.
repeated measures ANOVA on relative gap duration by tempo condition. The
results showed no eect of tempo (F (2, 34) = 1.7, p = 0.198). The relative
gap duration was in the range of 2529% for all three tempo conditions (slow
tempo: 0.29; medium tempo: 0.27; fast tempo: 0.25). This suggests that in
portato articulation the tongue-reed contact duration remains constant, inde-
pendent of the playing speed, while in staccato articulation, the relative gap
duration is constant.
Inuences of the measurement setup to the performances

In the questionnaire, we asked how comfortable the participants felt while
playing the sensor-equipped saxophone reed. The reed quality had to be rated
between 1 (very good) and 7 (very bad). Results showed that the reed quality
was evaluated as medium quality (M = 3, SD = 1.5). Participants also had to
indicate whether they felt comfortable when playing the sensor instrument or
not. We tested timing accuracy and timing precision for this group eect by
separate between-subjects ANOVAs. We found no signicant eect between
the two groups, thus, the sensor instrument did not aect the recorded perfor-
mances. We also tested for eects of self-reported handedness and skill level
(years of playing the instrument), but found no signicant eects.
3.3 Experiment 2: Listening test
We were interested in the abilities of listeners with dierent expertise in music
performance to discriminate between common articulation techniques in the
sound of the saxophone. Furthermore we were interested whether motor ex-
pertise in saxophone performance would facilitate the perception of saxophone
articulation.
3.3.1 Method
Participants
Nineteen female and twelve male (N = 31, mean age = 24 years, range = 19
32 years) students from Vienna music conservatories and Vienna universities
participated in the listening study. The group consisted of ten saxophone
players, ten musicians that play an instrument other than the saxophone and
eleven non-musicians. The saxophone players had a mean of 10.5 years (range
= 515 years, SD = 3.06) of experience in playing their instrument: eight of
them also participated in the production experiment described above (Section
3.2). The group of musicians, who did not play a wind instrument, had a
mean of 15.3 years (range = 822 years, SD = 3.83) of experience in musical
practice of various instruments. The group of non-musicians were students
of other elds, but 8 subjects had musical training in their early childhood,
with a mean of 4.5 years. Only one of the non-musicians had experience with
playing a wind instrument (the recorder).
Experimental Design
In a 3 × 3 × 2 × 2 × 2 (3 articulations × 3 intervals × 2 registers × 2
players × 2 listening blocks) design, we tested which articulation techniques
our participants were able to discriminate. We recorded note transitions of
three dierent pitch intervals (major second, minor sixth, minor sixth includ-
ing register change), with legato, portato and staccato articulation, within
two registers by two dierent players (one of them also participated in the
production experiment, the other is the rst author of this paper) on the same
alto-saxophone (YAS 32, by Yamaha Corp.), using their mouthpieces (AL3,
by Vandoren; Original 7*3, by Claude Lakey). Recordings were made in an
anechoic chamber using a microphone (C414, by AKG Acoustics) and Labview
hardware and software (DAQ LabView 2011, by National Instruments Corp.)
for recording the stimuli (44.1 kHz sampling rate, 16 bit resolution). Both play-
ers used synthetic sensor equipped saxophone reeds (Section 3.2.1), to ensure
tongue-reed contact in the portato and the staccato playing conditions. Dur-
ing the recordings, both players heard a metronome click on headphones (108
bpm for larger intervals, 120 bpm for small intervals), to produce consistent
timing in the stimuli. We recorded three eighth-notes (two note transitions,
see Figure 3.3) for each audio le. The beginning and the ending of the audio
le was edited with a volume fade-in and fade out, to limit the sound of ar-
ticulatory actions to only the two note-transitions. In total, 36 dierent audio
les were comprised in the pool of stimuli.
Stimuli were presented to the participants in the form of an ABX listening
test on a laptop computer. A java-based software program enabled the partic-
ipants to click on one of 3 buttons (A-Button, B-Button, X-Button) to play
back one stimulus. Buttons A and B contained two note transitions played
with dierent articulation techniques. Button X contained a third recording
that matched the articulation used in either A or B. The question our par-
ticipants had to answer was: Does X sound like A or B ? Listeners had to
decide whether X was more similar to A or B. Responses and reaction time
were recorded by the software.
Procedure
The experimental procedure complied with the Declaration of Helsinki: Par-
ticipants gave written consent prior to the experiment. All participants worked
on the same laptop computer (by ASUSTeK Computer Inc.) in a quiet en-
vironment and used the same studio headphones (K121, by AKG Acoustics).
They could adjust the playback volume to a comfortable level. Each par-
ticipant had 5 practice trials to learn how to navigate the ABX listening test
software. A pop-up on the screen indicated when the actual experiment began.
The experiment was grouped into two blocks, each containing all 36 stimuli
in four dierent listening orders. Participants were allowed to play back the
stimuli as often as required to make an assured judgement. After the rst
block was done the participants lled in a questionnaire about their musical
background. Afterwards the participants made another set of judgements. The
entire experiment lasted for about 30 minutes per participant.
3.3.2 Results and discussion

Overall, participants from all three groups were able to accomplish the listening
test with over 87% of correct answers. A Chi-squared test revealed no signif-
icant dierence on correct answers between the two repeated listening blocks
[χ2 (1) = 2.59, p = 0.11]. No eects of listening order [χ2 (3) = 5.01, p = 0.17]
or recording saxophonist [χ2 (1) = 0.05, p = 0.82] were found either. Due to a
labelling mistake in the playback list of stimuli, one stimulus pair had to be
1
excluded from the results.
To convert dichotomous response data to an interval-scale level, we com-
puted the percentage of wrong answers per participant collapsing across listen-
ing blocks and players. A two-way ANOVA on percentage of wrong answers,
with articulation (type of articulation to discriminate) as within-subjects and
listeners expertise as between-subjects revealed a signicant eect of the ar-
ticulation [F (2, 56) = 187.825, p < 0.001, η 2 = 0.933] and a signicant eect
of the listener's expertise [F (2, 28) = 4.167, p < 0.05, η 2 = 0.479], as well
2
as a signicant interaction [F (4, 56) = 5.847, p < 0.001, η = 0.543; see Fig-
ure 3.8a]. A two-way ANOVA on the response duration with articulation as
within-subjects and listeners expertise as between-subjects factor revealed a
signicant eect of articulation (F (2, 56) = 67.897, p < 0.001, η 2 = 0.841), but
no signicant eect of expertise or interactions (see Figure 3.8b).
Focussing on articulation, post-hoc pairwise t-tests showed that results from
the legato-portato listening task diered signicantly from the results of the
other two tasks (p < .001). Errors occurred most often when participants
had to discriminate between legato articulation and portato articulation (25%
1 A, B and X stimuli contained three dierent articulations: no correct answer was pos-
sible.
a) b)
35
20
● Musicians ●
● Musicians
● ● Non−Musicians ● Non−Musicians
30
Saxophonists Saxophonists
●
15
Wrong answers (%)
25
Response time (s)

●
20
10
●
15
●
●
●
10
5
5
● ●
0
0
legato−portato legato−staccato portato−staccato legato−portato legato−staccato portato−staccato
Articulation Articulation
Figure 3.8: Results of listening experiment: a) percentage of mistakes for discrimi-

nation task; (b) response time to accomplish the task; Participants were grouped by
their expertise in music making. Error bars show the standard error of the mean.
wrong answers, compared to < 1% wrong answers for remaining articulation
types). Additionally, Figure 3.8b shows the highest response durations for the
legato-portato condition.
Concering the listeners expertise, Figure 3.8 shows that non-musicians gave
more wrong answers (32%) and required more time than the other two groups
to respond (duration to answer per question M = 19.35 s, SD = 19.5), followed

by non-wind-instrument players (23% wrong answers; M = 15.79 s, SD =
13.8) and saxophonists (18% wrong answers; M = 11.13 s, SD = 8.7). The
results from our listening test suggest that musical expertise alters the abil-
ity to discriminate subtle sound dierences, like between legato and portato
tone transitions. The distinct sound of staccato tone transitions was well dis-
criminated from the other two articulation techniques, by all three groups of
listeners.
3.4 General discussion
This study investigated the production and the perception of articulation on
the saxophone with two experiments. For the production experiment we built
a sensor equipped saxophone reed to monitor tongue-reed interaction in alto-
saxophone performance, while participants performed melodies at three tempi
with dierent articulation techniques. The captured sensor-reed signals showed
that for portato articulation, the tongue-reed contact duration was indepen-
dent from the given tempo, whereas for staccato articulation the gap between
the tones was relative to the given tempo. In legato articulation, no tongue
strokes occurred and tone transitions were initiated by a change of the nger-
ings. Such coordination tasks occur with all wind instruments, where dierent
eectors (tongue and ngers) are required to produce one tone (e.g., ute,
clarinet, trumpet). It is also the case for string instruments that the player
has to coordinate dierent eectors to produce one tone. Bowing movements
with the right arm have to be coordinated with left hand ngerings. Baader
et al. (2005) looked at bow-nger coordination in violin playing and recorded
tone sequences where subjects had to play a sequence of tones, in which each
tone was initiated with a bow stroke and a nger change. The focus of their
study was primarily on bow-nger synchronization, which was shown to be far
from perfect simultaneity (50 ms), but did not lead to audible interruptions.
To play larger intervals on woodwind instruments the player has to close or
open multiple tone holes at the same time. This requires simultaneous nger
movements, also called safe nger transitions. Almeida et al. (2009) showed
that for ute performance unsafe nger transitions with approximately 21 ms
already lead to audible changes in the radiated sound. Taking these studies
into account, it seems that wind instrumentalists need even more precise nger
movements, which additionally have to be coordinated with the movements of
tongue.
In our study we looked into temporal eects of saxophone performances
under dierent tempi, which were produced by dierent eectors (ngers only,
tongue only, combined tongue-nger actions). We found that at the slow
tempo, tone onsets produced by tongue-only actions were signicantly less
precise than tone onsets produced by ngerings only. Highest precision was
achieved for combined tongue-nger actions. This corresponds to our hypoth-
esis that timing precision improves for combined tongue-nger actions. How-
ever, we did not expect to see that in the fast tempo condition, tonguing alone
was more precise than nger-only actions and combined tongue-nger actions
showed a high timing variability, at approximately the same level as nger-only
actions. This nding suggests that ngers play a dominant role in the overall
timing of saxophone performances.
In woodwind performance nger actions usually do not receive the same
attention as with piano playing, where the nger movements directly produce
the sound. Our observation that there is a strong inuence of nger timing on
the overall timing in woodwind performance may put a new focus on further
investigations of nger movements, nger trajectories, and nger forces in this
domain. With the help of sensor-equipped wind instruments and the devel-
opment of new customised sensors, useful advice for music education may be
gained in future research.
In the listening experiment, we observed that the articulatory sound modi-
cations (legato, portato, staccato) were mostly perceivable for non-musicians,
musicians (not playing saxophone) and professional saxophonists. Only the
sound of portato tongue-reed strokes was dicult to discriminate from that of
non-tongued legato tone-transitions. There are two possible reasons for this.
First, a brief damping of the reed vibrations does not immediately stop the
standing wave in the resonator and thus only slightly modies the radiated
sound. Not all listeners notice that the reed has been stopped. Second, unsafe
nger transitions in legato playing may also cause small gaps in the sound,
which non-experts may confound with portato tonguing (Almeida et al., 2009).
Nevertheless, the group of professional saxophonists was superior in discrim-
inating legato from portato sounds. An interesting observation was that the
one participant, who did not use any tonguing during the production exper-
iment and was therefore excluded from analysis there, also showed the worst
results in the listening experiment (in the group of professional saxophone
players). This strengthens our assumption that expertise in the underlying
motor-actions to modify sound, facilitates perceptional discrimination of such
sound modications. This conclusion is in line with the motor theory of speech
perception (Galantucci et al., 2006): the link between perception and produc-
tion of speech may also apply for the perception of articulation in saxophone
music performance. As a consequence, learning to play a musical instrument
enhances the ability to perceive more details of musical performance on that
instrument.

Chapter 4
Finger forces in clarinet playing
This chapter reports a study in which nger forces of 23 skilled clarinetists
were investigated during performance. The participants played eight carefully
selected excerpts from the rst Weber Concerto and a technical exercise (rep-
etition of the saxophone sound production task from Chapter 3) on a sensor
equipped Viennese clarinet.
The clarinet was developed at the Vienna Technical University within the
FWF funded project Measurement and analysis of nger forces in clarinet
playing. Special ring-shaped force sensors were developed during the project
and they were attached to six tone holes on the instrument. Details about
the construction of the low temperature co-red ceramic (LTCC) sensors can
be found in the related Ph.D. thesis by Weilguni (2013). For the experiment
presented in this chapter, the sensor clarinet was used for the very rst time in
an empirical study. This study not only discusses the measured nger forces in
clarinet playing, it also evaluates the unique nger force measurement setup.
4.1 Introduction
Clarinetists use their ngers to close or open the tone holes of the instru-
ment during music performance. Finger actions at the tone holes are used to
modulate the sounding frequency, thus quick and precise nger movements are
required for professional performance (Almeida et al., 2009). Clarinetists prac-
tice for decades until they gain a professional level. Besides studying musical
repertoire, practising involves technical training including scales, arpeggios and
67
68 Chapter 4. Finger forces in clarinet playing
other exercises to learn how to make well-controlled tone transitions (Mauz,
2011; Michaels, 1999).
Holding a musical instrument in a proper way during playing is dicult
and has to be learned. In contrast to lifting an arbitrary object (mostly with
thumb and index nger, Forssberg et al., 1991), a musician has to hold the
instrument in a certain position to play it. Especially with woodwind instru-
ments, this often involves lifting the instrument weight, but remaining exible
with the index-nger, middle-nger, ring-nger and pinkie nger of both hands
to operate the tone holes. Holding a clarinet requires balancing the instrument
on the thumbs without gripping it with the other ngers. Imbalanced stress to
the thumbs has already been identied to be a reason for overuse syndromes
with clarinetists (Diethelm, 2011).
Overuse syndromes are a general problem for professional music performers
across various instruments (Tubiana et al., 2005; Altenmüller and Jabusch,
2004). Consequently it is of interest for musicians, music teachers and medi-
cal personnel to gain insight into force related player-instrument interactions
(Grosshauser and Troester, 2013; Kinoshita et al., 2012; Guillemain et al.,
2010; Almeida et al., 2013; Bertsch and Mayer, 2005).
With string instruments the ngertip force to the strings has been a focus
of several studies. Kinoshita and Obata (2009) designed a violin with a force
transducer under the ngerboard to measure string clamping force of the left
hand. The average force peak was measured with 4.5 N, and they reported
decreasing ngertip forces (1.7 N) for faster tempi. Another aspect of their
investigations was the eect of varying dynamics. Although the dynamics
in violin playing are controlled by right hand bowing techniques, there was
a signicant inuence of the performed dynamics on the left-hand ngertip
force. Louder playing showed increased left-hand ngertip forces. In another
study, Kinoshita et al. (2012) compared expert players with novice players
and found that the experts' clamping on the strings was signicantly larger.
A reason might be that holding the string tight to the ngerboard is required
to produce a good sound quality and helps to avoid undesired sounds from the
strings hitting the nger board. Hori et al. (2013) measured ngertip forces
for holding down the strings of the guitar with about 3050 N. An analysis of
the resulting sound showed that larger force (pressing down with around 50
Chapter 4. Finger forces in clarinet playing 69
N) improved the tone quality and he concluded that this is the proper holding
technique.
However, in woodwind instruments the applied nger force to the tone hole
does not inuence the resulting sound. Consequently it can be assumed that
the player has to use barely minimal nger forces to close the tone holes air
tight. This is also recommended by clarinet teachers (Wehle, 2007).
Investigations of ancillary gestures in expressive clarinet performance have
shown that a lot of body movements happened, that were not related to eco-
nomic control over the instrument only (Desmet et al., 2012; Caramiaux et al.,
2012). Circular movements of the clarinet bell, head and shoulder movements
of the performer, bending at the waist or with the knees, apping the arms
and stepping with the feet were typical ancillary gestures. Wanderley et al.
(2005) also observed that for each performer similar movements appeared for
repetitions of the same piece. Palmer et al. (2009a) explained further that an-
cillary gestures in clarinet playing were related to acoustic features of musical
expression. Such an embodiment to the musical expression might also aect
the nger force proles during expressive music performance. Furthermore, to
play with loud dynamics or in the high register, the player has to blow with
more pressure into the instrument. Does such a tension of the respiratory
system inuence the nger forces applied to the tone holes?
Motion capture studies focusing on the ngerings of clarinetists reported
that during dicult and fast sequences, players used larger nger movements
and increased accelerations at nger-key contact (Palmer et al., 2007, 2009b).
Palmer et al. (2009b) suggest that there is an eect of tactile feedback on tem-
poral accuracy. Increased sensory information to the performers ngers seemed
to improve the timing accuracy of the played sequence. Similar observations
for ute performance were made and may support this hypothesis (Almeida
et al., 2009).
Hofmann and Goebl (2014, see Chapter 3) measured the eect of dierent
eectors used (tongue-only actions, nger-only actions, combined tongue-nger
actions) to play melodies on the alto-saxophone. The observations made have
partly supported the multiple-timer model by Ivry et al. (2002). In the case
of the slow tempo condition, combined tongue-nger actions stabilized the
overall timing. For fast tempi, nger actions showed a dominant inuence on
the performance timing. Furthermore, in that experiment two melodies were
used as stimuli. One melody focused on pressing the keys of the instrument
and the other focussed on releasing the keys. No inuence of direction of nger
movements on the performed timing was found, similar to reports from other
studies Palmer et al. (2009b). Consequently, a repetition of this experiment,
does not require to measure both directions of nger movements.
However, a repetition of this study on a dierent single-reed instrument,
would allow to verify the observations made with saxophonists. Moreover, it
would be interesting to proof if a clarinetists, with a classical music background
show similar results as the group of jazz and classical saxophonists. With
a nger force sensor clarinet it will additionally be possible to measure the
applied force to the tone holes and report nger forces used with this technical
exercise, requiring actions close to the synchronization threshold of professional
musicians (Repp, 2005).
The foreseen study focusses on measuring the nger forces of index nger,
middle nger, and ring nger of both hands during expressive clarinet per-
formance and will also include a repetition of the technical exercise from the
saxophone study. A special Viennese Clarinet, equipped with six force sensor
rings at the tone holes of the instrument (Weilguni, 2013) together with strain
gauge sensor equipped single-reeds (Hofmann et al., 2013a) will be used to
capture nger force proles and tongue actions of professional clarinetists.
4.2 Performance experiment
4.2.1 Methods
Participants
Ten female and thirteen male clarinetists (N = 23, mean age = 26.9 years,
range 1945 years) participated in this study. On average the participants
played the clarinet for 17.26 years (range 937 years) and reported to practice
with their instrument for about 2.78 hours per day (SD = 1.16). Seventeen
participants were students from the University of Music and Performing Arts
Vienna and six participants were professional performers, ve of them were
additionally teaching at an academic institution.

Experimental design
For the experiment three dierent performance tasks were prepared: a warm-
up task, an expressive performance task, and a technical exercise task.
For the warm-up task, a simple melody was designed with a focus on se-
quentially opening and closing the six sensor equipped tone holes (nd score
in Appendix B.1).
For the expressive performance task, excerpts from concert clarinet litera-
ture were chosen to measure nger forces under dierent musical situations. In
a 2 × 2 × 2 × 2 design (register: lowhigh; tempo: slowfast, dynamics: soft
loud, expression level: lowhigh), nger forces during clarinet performances
will be investigated. Eight excerpts from the Clarinet Concerto No. 1 in F
minor (Op. 73) for clarinet in Bb, by C.M.v.Weber were selected (nd score in
Appendix B.2 and B.3). This particular piece was chosen, because it belongs
to the standard repertoire of clarinet students at the University of Music and
Performing Arts Vienna. Each selection fullled one of three testing condi-
tions (see Appendix B.1), the fourth testing condition will be introduced by
the experimenter during the experiment.
For the technical exercise task, an isochronous 23-tone melody was designed
for the experiment (Appendix B.4), similar to the melody used in the saxo-
phone study (see Chapter 3 or Hofmann and Goebl, 2014): The rst eight
notes of the melody are a repetition of the same tone. The notes after (923)
are performed by closing the tone holes with the ngers of the left hand. The
score was prepared for clarinet in Bb and included the articulation instruc-
tions for portato, staccato and legato playing. Because of the reason that note
repetitions in legato are not possible to play, the notes 18 had to be removed
from the score for this playing condition.
Equipment
The following setup was chosen for this experiment: a sensor-equipped Vien-
nese clarinet in Bb (see Figure 4.2), a microphone, a digital metronome, a small
camera, and a multi-channel recording device. To measure the reed vibrations
and the tongue actions of the players during performance, sensor reeds were
prepared for this study (see Chapter 2.2 for details). Six ring-shaped force sen-
Figure 4.1: Ring-shaped sensor attached to a tone-hole of the clarinet: (left) without
polymer ring; (right) with polymer ring glued on three soda lime glass balls with epoxy
adhesive. The surrounded key-work has been dismounted for this picture (Photos by
Michael Weilguni).
Figure 4.2: Viennese Sensor Clarinet in Bb (Model38, by Foag), with six force sensor
rings mounted to the tone holes and a sensor clarinet reed.
sors were mounted around the six front tone-holes of the Bb clarinet (Model
38, by Martin Foag, Hafenhofen), to measure the nger forces of index-nger,
middle-nger, and ring-nger of both hands during clarinet performance (Weil-
guni, 2013). The sensor reed, the force sensor rings, the microphone (d:vote
4099, by DPA Microphones) and the digital metronome (KDM-1, by Korg
Inc.) were connected via BNC cables to a multichannel analog-digital con-
verter (DAQ LabView 2011, by National Instruments Corp.) to record the
signals simultaneously (A/D conversion with sampling rate 11.025 kHz, 16 bit
resolution). The data was captured onto computer hard disk .

Procedure
The experiment was conducted in accordance with the Declaration of Helsinki.
All participants gave their informed consent to the procedures of the study and
received a nominal fee after the experiment.
In the beginning of the experiment each participant chose a sensor reed
according to the preferred reed strength (Légère Nick, Bb clarinet, german cut,
reed strength: 2.54.5). Participants were allowed to use their own mouthpiece
but played on the same sensor equipped clarinet in Bb. Before the recordings
started, participants had ve minutes time to adjust the setup according to
their needs.
The main experiment contained ve tasks: First, the warm-up task was
performed to a metronome signal with 120 beats per minute on each quarter-
note beat Appendix B.1. Second, the expressive performance task, containing
eight excerpts from the Weber Clarinet Concerto, was played. We recorded
four trials for each excerpt. The instructions for the rst two trials were, to
focus on technical aspects of playing (e.g., correct tones, synchronisation to
the provided metronome signal). For the remaining two trials, the metronome
signal was muted and the instructions were to play expressively (choose own
tempo, focus on dynamics), like in a concert situation. Third, participants
lled in a questionnaire about their musical background and gave feedback
about playing the sensor equipped clarinet.
For a technical exercise (fourth) task, each participant played the 23-tone
melody (Appendix B.4) in legato, portato and staccato articulation in three
tempo conditions (slow, IOI for eighth notes = 250 ms; medium, IOI = 178.6 ms;
fast, IOI = 144.2 ms). The metronome provided the synchronisation signal on
a quarter-note level. For the rst two repetitions, participants synchronised
with the metronome. When the metronome was muted, they continued play-
ing, until the melody was played six times in total. We recorded two trials per
tempo condition, which were ordered always from the slowest to the fastest
(to repeat the procedure from Hofmann and Goebl, 2014). In a nal, fth
step, the warm-up task was performed again to the metronome signal. The
experiment lasted for approximately 1 hour per participant.

Data analysis
Each of the six force sensor rings attached to the tone holes of the sensor clar-
inet, contained three measurement cells with 120◦ distance (see Figure 4.1).
Calibration values for the measurement cells were provided by the manufac-
turers from the TU Vienna (Weilguni, 2013), to convert the measured output
voltage to force in Newton (N).
During the experiment it turned out that the sensors were fragile. For exam-
ple sensor 4 (right-hand index nger) broke completely through the mechanical
burdening while recording the sixth participant. All three measurement cells
from this sensor fell o. The measurement cells were glued with conductive sil-
ver/epoxy adhesive, into the sensor ring. Even putting them back in changed
the behaviour of the sensor. Figure 4.3 (bottom) shows that the force values
captured by Sensor 4 (dotted line), are comparable with the measurements
from the other sensors until player 6, and than for player 79 it produces
wrong data. For player 1013 and 1723 the sensor did not capture any data
at all. Consequently all measurements from sensor 4 were removed from fur-
ther analysis steps. Unfortunately, small artefacts occurred in the data of all
sensors at some point. A reason for this were the cable connections between
the upper and the lower part of the clarinet, going over the right hand thumb
of the player. If the player touched the cables, this inferred the data. Fur-
thermore, the sensors showed the tendency to drift during the study. The 0 N
had to be calibrated for each recording session individually, and in some cases,
they also drifted during one experiment. As a consequence, all the captured
data of each measurement cell had to be carefully inspected. If artefacts or
such a drift occurred, the measurement cell was excluded from the analysis.
The values of each working measurement cell from one sensor ring were
summed to calculate the force applied by each nger to the sensor ring. Ex-
amples for nger force proles are shown in Appendix B.6 and B.7.
If data from one or two measurement cells was excluded, weighted sums
were used to compensate this. When all three measurement cells produced
wrong data, no nger force prole for this sensor ring exists. Figure 4.3 shows,
that at the beginning of the experiment (ID 1) all 6 sensors worked properly,
and for the last participant (ID 23) only 3 sensors produced reliable data.
From the nger force proles, the regions where the player closed the tone
hole were selected automatically by using a threshold. For the selected regions,
the peak force (Fmax ) and the average force (Fmean ) were calculated and the
required force to press down the key-work around the tone hole was added.
Post processed nger force measurement results were saved per nger and per
recorded trial for further statistical analysis.
To examine timing properties of the technical exercise task, tongue-reed
contact (TRC) and tongue-reed release (TRR) landmarks were extracted from
the captured reed signals. The landmark detection function and the score-
performance matching were similar to the method reported for the saxophone
study in Chapter 3.2.1 (p. 49).

Mean Finger Forces per Participant
76
5
Ring Sensors
● 1
4
● 2
3
3
●
●
4
● ● ●
●
5
●
6
2
● ● ●
● ● ●
● ●
● ● ● ● ●
● ● ● ● ●
● ● ● ●
●
1
● ● ● ●
Mean Finger Force (N)

● ● ●
● ●
● ●
● ● ●
●
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
Participant ID
Peak Finger Forces per Participant
15
Ring Sensors
● 1
● 2
10
3
● 4
5
●
6
●
5
● ●
● ● ● ● ●
●
● ● ● ●
● ● ●
● ●
Peak Finger Force (N)

● ● ● ● ● ● ● ●
● ● ●
● ● ● ● ●
● ● ● ●
● ● ● ●
●
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
Participant ID
Figure 4.3: Top: Average nger forces (Fmean in N) measured for each participant during the expressive performance task (excerpts from
Clarinet Concerto No. 1 by Weber). Each line represents the force data captured for one nger during the entire task (1 = left-hand
index nger, 2 = l-h. middle nger, 3 = l-h. ring nger, 4 = right-hand index nger, 5 = r-h. middle nger, 6 = r-h. ring nger).
Bottom: Average peak forces (Fmax in N) measured for each participant.
Chapter 4. Finger forces in clarinet playing
4.2.2 Results
Finger forces during expressive performance task
The average nger force measured for all participants during the expressive
performance task (excerpts from Weber Concerto No. 1) was Fmean = 1.17
Newton (N) with a standard deviation of SD = 0.37. Measured peak nger
forces (Fmax ) over all recorded trails varied between 0.84 N and 12.95 N with
an average of 3.05 N and a standard deviation of 2.0 N. Figure 4.3 (top) shows
the average measured nger force for each participant's nger operating on a
sensor equipped tone hole. Each line represents the captured data of one force
sensor ring during the experiment.
Looking into the average peak nger forces, it is obvious that participant
4 showed the highest peak nger forces captured during the expressive perfor-
mance task (above 12 N). To compare individual nger force proles, Appendix
B.6 (ID 2) and Appendix B.7 (ID 4) depict force measurements for excerpt A
1
of the expressive performance task . Note in Appendix B.5 (right) that par-
ticipant 4's ngertips already changed its color when closing the tone holes.
A one-way analysis of variance (ANOVA) on Fmean with participants' sex
as between-subjects factor revealed no signicant eect of the player's sex
[F(1,21) = 0.703, p = 0.411], although females average nger forces were slightly
lower than the nger forces of the male participants (female Fmean = 1.1 N;
male Fmean = 1.2 N). Taking the professional level of the players (students vs.
professionals) into account, showed also no signicant eect with an one-way
ANOVA on Fmean testing participants' professional level as between-subjects
factor [F(1,21) = 1.081, p = 0.310].

To examine nger forces under dierent situations of musical expression,
participants played eight excerpts with dierent musical instructions given by
the score (register: lowhigh; tempo: slowfast, dynamics: softloud). Each
sequence was performed under two dierent playing instructions from the ex-
perimenter. In one condition, the clarinetists had to focus on technical aspects
of playing (e.g., correct pitch and precise note length, synchronisation to the
1 Raw data of left-hand index nger force proles, captured for all participants performing
both trials of excerpt A in the highly expressive playing condition are plotted in Appendix
B.8, B.9, and B.10.
Expression
1.5
1.4
1.3
1.2
●
●
1.1
1.0
0.9
Low High
Expression Level
Figure 4.4: Mean nger forces (Fmean in N) for expressive performance task, in-
creased with a higher level of musical expression. Error bars show the standard error
of the mean.
metronome), where in the second condition the clarinetists were instructed to
play with a high level of musical expression, similar to a concert performance
in front of an audience (expression level: lowhigh).
A one-way repeated measures ANOVA on Fmean by expression level indi-
cated a signicant main eect of expression level during performance [F(1,22) =

26.2, p < 0.001, η 2 = 0.74]. Figure 4.4 shows that force values increased for
performances with a higher level of expression.
To examine nger forces how they would appear in performance situations,
only the trials recorded in the high expression level performance condition are
further analysed and discussed. A three-way repeated measures ANOVA on
Fmean by register, dynamics, and tempo revealed three signicant main eects
(register [F(1,22) = 19.04, p < 0.001, η 2 = 0.681]; dynamics [F(1,22) = 4.56, p <
0.05, η 2 = 0.414]; tempo [F(1,22) = 5.22, p < 0.05, η 2 = 0.438; see Figure 4.5])
and two interactions (register and tempo [F(1,22) = 18.39, p < 0.001]; dynamics
and tempo [F(1,22) = 23.52, p < 0.001; see Figure 4.6]). Playing in the high
register led to increased nger forces, as well as louder dynamics (Figure 4.5
a, b). Overall, sequences that required playing at fast tempi showed reduced
nger force values compared to the slower parts (Figure 4.5c). Taking the
a) Register b) Dynamics
1.5
1.5
1.4
1.4

1.3
1.3
●
●
1.2
1.2
●
●
1.1
1.1
1.0
1.0
0.9
0.9
Low High Soft Loud
Register condition Dynamic condition
c) Tempo
1.5
1.4
1.3
●
1.2
●
1.1
1.0
0.9
Slow Fast
Tempo condition
Figure 4.5: Mean nger forces (Fmean ) for dierent musical instructions. a) Playing
in high register led to increased mean nger forces. b) Increasing dynamics increased
nger forces. c) Playing in a fast tempo resulted in reduced nger forces. Error bars
show the standard error of the mean.
a) Register and Tempo b) Dynamics and Tempo

1.8
1.8
● Slow Tempo ● Slow Tempo
● Fast tempo ● Fast tempo
1.6
1.6

1.4
1.4
●
●
●
1.2
1.2
●
●
●
● ●
1.0
1.0
Low High Soft Loud
Register condition Dynamic condition
Figure 4.6: Interactions between register, dynamics and tempo of performances. a)

In slow tempo, participants' nger forces increased signicantly when playing in high
register but were nearly constant for fast tempi (solid line). b) Loud dynamics led
to increased nger forces in slow tempo, but showed the opposite behaviour in fast
tempo. Error bars show the standard error of the mean.
signicant interactions into account, separate post-hoc t -tests, two for each
tempo condition were necessary. The result indicates that in the slow tempo
condition, register (p < 0.001) and dynamics (p < 0.001) had a signicant
inuence on nger force in terms that increased dynamics and playing in high
register showed increasing nger forces (Figure 4.6a, b; dashed line). Whereas
in the fast tempo condition only dynamics (p < 0.05) signicantly inuenced
the nger forces, and register had no further inuence (p = 0.69; Figure 4.6a,
solid black line). It is interesting to note in Figure 4.6b that depending on the
tempo the direction of inuence through dynamics changed. In slow tempo
(dashed line), loud dynamics led to increased nger forces, whereas in fast
tempo (solid line), nger forces decreased with louder dynamics.
Self-evaluation of nger forces

In the questionnaire, participants were asked to self-evaluate their nger forces,
without knowing any results of the measurement. Participants had to report on
a seven-step rating scale (from 3 to +3), whether they think their own nger
2.5
●
Measured Mean Finger Force (N)
2.0
●
1.5
● ●
●
● ● ●
● ●
● ●
●
1.0
●
●
●
●
●
●
●
0.5
0.0
−2 −1 0 1 2
Self Reported Finger Force
Figure 4.7: Self-estimation of nger forces (from 3 to +3) compared to measured

Fmean nger forces.
force is below average (from 3), on average (0), or on a higher force level
(up to +3). We correlated the measured nger forces (Fmean ) for expressive
performance with the answers of the questionnaire. The measured nger force
showed a high correlation with the players' self-evaluation of their nger force
[r = .64, p < 0.001; Figure 4.7].
We also asked the participants to report any physical discomfort in rela-
tion to clarinet playing. Non of the participants reported permanent physical
discomfort. For the participants who reported only occasionally discomfort
during playing (19%) or during and after playing (24%), an one-way repeated
measures ANOVA on Fmean with discomfort as between-subjects factor did not

reach a level of signicance [F (1, 19) = 3.62, p = 0.07, η = 0.4], but showed a
medium eect. However, participants who reported physical discomfort, self-
evaluated their nger forces as above average (+1.2) and also showed larger
nger force values during the experiment (Fmean = 1.4N ) compared to those
reporting no discomfort at all (Fmean = 1.08N ).

When asking the participants if they would be interested in using a force
sensor clarinet for practising or for teaching, the results showed no clear trend
(self-use: yes = 43 %; teaching: yes = 48 %). However, looking at the two
groups of players separately (reported physical discomfort or not) showed that

the group of players with problems had more interest (70%) in using such a
measurement device than the other group (27%).
Timing of the performed melody in the technical exercise task1

To examine the timing in the technical exercise task, the inter-onset inter-
vals between subsequent TRR (oneset) landmarks in the reed signal were cal-
culated (IOI x = tx+1 − tx ). From these IOIs, the timing error ((IOI obs −
IOI exp )/IOI exp ) and the coecient of variation (CV = SDIOI /M eanIOI ) was
derived (similar to the analysis in Hofmann and Goebl, 2014). The timing
error (timing accuracy) is a measure for the deviation from the given tempo,
where negative values indicate a sequence played too fast and positive values
correspond to a sequence played too slow. The coecient of variation (CV,
timing precision) depicts the regularity of the tone onsets.
Figure 4.8a shows the timing error for all performed melodies, grouped by
synchronisation condition for all three tempi. The solid black line shows that
2
all clarinetists were able to perform the melodies to the metronome click .
Muting the metronome click resulted in participants playing too fast in the
slow tempo condition and too slow in the fast tempo condition (dashed line).
A three-way repeated measures ANOVA on timing error by tempo condition
(slow, medium, fast) and synchronisation condition (with metronome click =
synchronisation, without metronome click = continuation) as within-subjects
and participants' professional level (students vs. professionals) as between-
subjects factor conrmed a signicant main eect of tempo [F (2, 40) = 18.645,
p < 0.001, η 2 = 0.695] as well as a signicant interaction between tempo and
synchronisation [F (2, 40) = 17.093, p < 0.001, η 2 = 0.679] and tempo and
2
professional level [F (2, 40) = 6.232, p < 0.01, η = 0.487]. No main eect of
professional level on timing accuracy (p = 0.424) nor any other signicant
interactions were found.
The performed sequences showed a high timing precision (mean CV = 0.09).
The same three-way ANOVA calculated for the CV indicated only a signi-
1 In
this subsection, the same analysis methods as with the saxophone study (Hofmann
and Goebl, 2014) are used, in order to gain comparable data.
2 Data of one participant had to be omitted from the analysis, because the sensor data
was interfered by artefacts.

0.10
0.20
(too slow) ● Continuation ● Continuation
Mean signed timing error ● Synchronisation ● Synchronisation
0.05
0.15
Mean CV of timing
●
●
●
0.00
0.10
● ●
● ● ●
● ●
●
●
−0.05
0.05
−0.10
0.00
(too fast)
Figure 4.8: Timing error (a) and coecient of variation (b), for synchronization-
continuation playing conditions for the technical exercise. When playing with
metronome click (synchronisation phase, solid line) and without metronome click
(continuation phase, dashed line). Error bars show the standard error of the mean.
cant main eect of tempo [F (2, 40) = 3.611, p < 0.5, η 2 = 0.391]. Figure 4.8
shows that the regularity of subsequent events deteriorated with increasing
tempo. In comparison to the measurements on the saxophone (see Chapter
3.2.2), synchronisation condition did not reach a signicant level (p = 0.0584),

because the clarinetists' regularity of events was higher than with the saxo-
phonists, especially in the continuation condition. The timing quality (timing
error, timing precision) for both participant groups (students and profession-
als) was in general the same. A possible reason that no signicant eect in
timing quality between the two groups (students vs. professionals) was found,
might be that the group of professional players was small (N = 6) and the
participating students played their instrument on average for 13 years. All
students had successfully passed the music university entrance exams which
indicates that they already perform on a high level.

Timing with multiple eectors in the technical exercise task1

The melody of the technical exercise consisted of three parts. The rst part
required only nger actions (legato), the second part required only tongue
articulation (portato tone repetition) and the remaining notes required si-
2
multaneous tongue and nger actions . Onset timing of the three parts was
compared by analysing timing error and timing precision (Figure 4.9). A
two-way repeated measures ANOVA on the timing error by eector combi-
nation and tempo, showed a signicant main eect of the executing eector
[F (2, 42) = 8.95, p < 0.001, η 2 = 0.547], as well as a main eect of tempo
2
[F (2, 42) = 18.52, p < 0.001, η = 0.685] (no interactions). A post-hoc pairwise
t-test showed that timing with only nger actions was signicantly dierent
from only tongue actions (Bonferoni: p < 0.001) and combined tongue-nger
actions (Bonferoni: p < 0.001). Similarities to the results of the saxophone
experiment (see Chapter 3.2.2 or Hofmann and Goebl, 2014) were found in the
clarinet data: Playing with the ngers alone led to faster performances and
playing with the tongue alone led to a reduction of the tempo compared to the
metronome. Combining both eectors stabilized the timing error.
A two-way repeated measures ANOVA on temporal precision (CV) by eec-
tor combination and tempo, resulted in two signicant main eects but showed
no interactions. Figure 4.9b depicts how timing precision deteriorated with in-
creasing tempo [F (2, 42) = 20.18, p < 0.001, η 2 = 0.7] and was also aected
by the combination of eectors [F (2, 42) = 6.317, p < 0.01, η 2 = 0.481]. The
measured values for nger-only timing precision were on approximately the
same level as with the saxophone study and showed a similar tempo pattern
(clarinetists' mean CV = 0.097, Figure 4.9b; saxophonists' mean CV = 0.10,
Figure 3.6b).
A main dierence in the results of both studies was that the clarinetists'
tongue-only actions showed a much higher temporal precision, especially in
the slow and the medium tempo condition (clarinetists' mean CV = 0.074,
saxophonists' mean CV = 0.11). Consequently, nger-only actions appeared
to have the lowest timing precision compared to the other eector combina-
1 In
this subsection, the same analysis methods as with the saxophone study (Hofmann
and Goebl, 2014) are used, in order to gain comparable results.
2 For this analysis the data set was reduced to legato and portato performances only.
0.10
0.20
(too slow) ● Fingers only ● Fingers only
● Fingers + tongue ● Fingers + tongue
Tongue only Tongue only
0.05
0.15
Mean CV of timing
●
● ●
0.00
0.10
● ● ●
● ●
●
●
● ●
−0.05
0.05
−0.10
0.00
(too fast)
Figure 4.9: Timing error (a) and coecient of variation (b), grouped by eectors
used to produce tone onsets. Error bars show the standard error of the mean.
tions across all tested tempo conditions (Figure 4.9b, dotted line), although it
showed the same CV than with the saxophonists. However, similar to the ob-
servations on the saxophone was that combined tongue-nger actions improved
timing in the slow tempo condition. Furthermore, at fast tempo, combined
tongue-nger actions were closer to the precision of the ngers-only condition.
Separate post-hoc pairwise t-test, three for each tempo condition, conrmed
that at slow tempo, combined tongue-nger actions were signicantly dierent
from nger-only actions (Bonferoni: p < 0.01). This observation would sup-
port the multiple timer theory (Ivry et al., 2002), where a coupling of multiple
eectors leads to improved timing. However, in the medium and the fast tempo
condition there was no signicant dierence between combined tongue-nger
actions and nger-only actions (Bonferoni: p = 1.0) observable. Moreover, the
timing of the ngers overruled the timing of the tongue when both eectors
were used in a coordinated fashion. In line with the observations on the saxo-
phone, this indicates that nger movements play a dominant role in the overall
timing precision of clarinet performance.

Timing and nger forces in the technical exercise task
Looking at the measured left-hand nger force for the technical exercise task,
participants used lighter ngering on the sensors (Fmean = 0.64 N) compared

1
to the expressive performance task . A two-way ANOVA on Fmean by tempo
condition and professional level showed a main eect of tempo [F (2, 42) =
16.751, p < 0.001, η 2 = 0.675]. Professional level of the players did not reach
a level of signicance [F (1, 20) = 1.653, p = 0.213] but showed a small eect
2
on the used nger force (η = 0.276). Figure 4.10 depicts that the nger
forces of professional players were softer compared to those of the students,
especially in the slow and in the medium tempo. Separate post-hoc t-tests
(Bonferoni) for each tempo condition conrmed this observation for the slow
and the medium tempo (slow tempo: p < 0.05; medium tempo: p < 0.05;
fast tempo: p = 0.153). In comparison to the expressive performance task,
faster tempi for the technical exercise led to increased nger forces. A possible
reason might be that the technical exercise melody was be played at only one
(comfortable) dynamic level, without musical expression and in the low register
of the instrument.
Weilguni (2013) hypothesed in his Ph.D. thesis that too intense pressing
of the clarinet keys may result in unwanted uctuations in tone durations
and limitations in the maximum playing tempo. To test whether there is a
correlation between the applied nger force to the keys and the maximum
playing tempo, we assume that higher nger force must result in a positive
mean signed timing error of performances. For each tempo condition, the
mean timing error was correlated with the measured nger forces (Fmean ). No
signicant correlation was found in our data (slow tempo: r = −0.10, p =

0.499; medium tempo: r = 0.08, p = 0.590; fast tempo: r = 0.14, p = 0.369).
The same correlation for the coecient of variation was calculated to test the
hypothesis of unwanted uctuations in tone durations, by correlating timing
precision with nger forces (Fmean ). No such correlation was found in our data
(slow tempo: r = −0.04, p = 0.843; medium tempo: r = 0.09, p = 0.675; fast
tempo: r = −0.14, p = 0.548). From a comparison with the measurements of
1 To
put a focus on the ngering technique the dataset was restricted to the performances
with the nger-only condition.
Finger forces in technical excercise
1.0
● Students
● Professionals
0.8
●
●
●
●
0.6
●
●
0.4
0.2
0.0
Slow Medium Fast
Tempo condition
Figure 4.10: Average left-hand nger forces for technical exercise task, grouped by
the professional level of the participants. Professional players used less nger forces
than clarinet students.
the expressive performance task (Fmean = 1.2 N for performances with a high
expression level), it is obvious that the participants used less nger forces in
the technical exercise task already (Fmean = 0.64 N).
Figure 4.11 depicts the force measurements (top) for each participant. The
bottom panel allows to compare the timing precision between all participants
and it is visible that participant 14's timing precision was superior in compar-
ison to the others. From the questionnaire it was possible to determine that
1
this was the only participant who reported left-handedness . It is interesting
to see that the timing precision of the left-handed player was superior (CV =
0.053) in comparison to the average precision of all participants (CV mean =
0.098).
1
The reason why this technical exercise focussed on left-hand ngerings was through a
limitation in the number of sensors available during the pre-saxophone study. To collect
comparable data, the melody was not changed for the successive experiments. Besides,
sequences with left-hand nger actions exist in musical scores as well and have to be practised
and performed by every player.
Mean finger forces per participant in technical excercise task
88
2.5
Tempo condition
● Slow
2.0
● Medium
Fast
1.5
●
●
●
● ●
1.0
● ● ● ●
● ● ● ●
● ● ●
●
● ● ● ●
● ● ● ● ● ●
● ● ● ●
0.5
● ● ● ●
●
●

0.0
1 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
IDs
Timing accuracy per participant in technical excercise task
0.2
Tempo condition
● Slow
● Medium
0.1
●
Fast
● ●
● ● ● ●
● ● ●
● ● ● ●
●
0.0
● ●
● ● ●
● ● ● ● ● ● ●
● ● ● ● ● ●
● ● ● ● ●
● ●
●
−0.1

−0.2
1 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
IDs
Timing precision per participant in technical excercise task
0.5
Tempo condition
● Slow
0.4
● Medium
Fast
0.3
●
0.2
● ●
● ● ● ●
● ● ● ●
● ● ●
Mean CV of timing
●
0.1
● ● ● ● ● ● ● ● ●
●
● ● ● ● ● ● ● ● ●
● ●
● ● ● ●
● ●
0.0
1 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
IDs
Figure 4.11: Top: Average left-hand nger forces (Fmean ) measured for each participant during the technical exercise task, performed
Chapter 4. Finger forces in clarinet playing
under the legato articulation condition (Bar 2 and 3 of score in Figure B.4) in three dierent tempi. Timing error (middle panel) and
coecient of variation (bottom panel) for the same performances.
Eect of fatigue over the duration of the experiment (warm-up task)

In the beginning of the experiment and at the very end, participants performed
a simple melody (Appendix B.1) to the metronome click. To measure an ef-
fect of fatigue, a two-way ANOVA on Fmean with participants' professional
level (students vs. professionals) as between-subjects and pre-post condition
as within-subjects factor was calculated. The results revealed no signicant
eect of fatigue [F (1, 20) = 0.153, p = 0.7] and no signicant eect of profes-
1
sional level of the players [F (1, 20) = 0.443, p = 0.513], nor any interaction .
The answers of the background questionnaire showed that the participants
usually practise for a longer duration daily (M = 2.78 h, SD = 1.15) than the
experiment lasted (approximately 1 h).
Inuence of the sensor instrument to the measured nger forces

In the questionnaire, participants evaluated the quality of the sensor equipped
clarinet reed, the instrument, and also indicated whether they felt comfortable
with playing this setup or not. The reed quality and the instrument quality
had to be rated between +3 (very good) and 3 (very poor). Results showed,
that the reed was evaluated as medium quality (M = 0.17, SD = 1.75) and
the instrument as well (M = 0.74, SD = 1.1). We tested Fmean for the group
eect of setup evaluation (players felt comfortable vs. players who reported
issues with the setup) by separate between-subjects ANOVAs and found no
signicant eect.
4.3 Discussion
This study investigated nger force proles and temporal eects of professional
clarinetists and advanced clarinet students while performing on a sensor clar-
inet. The measurement setup captured nger forces of index nger, middle
nger and ring nger of both hands, applied to the six main tone holes, as well
as the reed oscillations during performance.
The average nger forces participants used to close the tone holes during
expressive performance was measured with Fmean = 1.21 N. Compared to nger

1 Data from two participants had to be removed, due to intermission of the sensor data.
1
forces reported for other instruments , this value is low. Even less force was
used, when the participants had to perform a technical exercise task focussing
on articulation techniques and timing (Fmean = 0.64 N). Even though, the
average peak nger force captured during the entire study (Fmax = 12.95 N) is
high in relation to the average nger force, it is still below the recommended
nger force of 50 N recommended for proper holding a guitar string (Hori
et al., 2013). Hence, clarinet playing requires very low forces of index nger,
middle nger and ring nger of both hands, compared to playing other musical
instruments.
Practising and performing a musical instrument over a long time is an uni-
lateral stress for the body. Using too much tension during playing may result
in overuse syndromes. Force measurements have been shown to be a useful
tool to pinpoint reasons for overstressed body parts (Grosshauser and Troester,
2013). Although index nger, middle nger and ring nger have not primarily
been reported as problematic areas of clarinetists (Diethelm, 2011), we hy-
pothesised that using too much tension with these ngers may be an indicator
for an overall uptightness during performance. Such an uptightness may lead
to physical discomfort in relation to clarinet playing. However, the group of
players who reported physical discomfort in relation to their clarinet play-
ing, did not show signicantly dierent nger force proles compared to the
group of players reporting no such problems. Consequently, measuring forces
of these particular six ngers provides no reliable indicator to forecast overuse
syndromes in clarinet playing. Additionally, when participants had to report
how much nger force they think they use, they showed good results in their
self-estimation.
Measuring nger forces on the clarinet still remains a complicated proce-
dure. Although the ring-shaped sensors were capable to measure nger forces
applied to the six tone holes of the instrument, these sensors turned out to be
fragile. Some sensors already broke during this study. Possible reasons might
be a sliding from side keys to the tone holes and movements of the arms, as
1 Holding down violin strings was measured with (Fmean >2.7 N, Kinoshita and Obata,
2009), guitar strings clamping force was (Fmean >30 N, Hori et al., 2013) and also the force
required to play a grand piano in forte dynamics was measured with (Fmean >5 N, Parlitz
et al., 1998).
reported in Wanderley et al. (2005). Such player actions cause a rotation of
the ngers. The occurring rotational forces work against the construction prin-
ciple of the ring shaped force sensors, which were designed to be pressed only
from a perpendicular direction. In worst case, rotating the sensor disengaged
the polymer ring from the glass balls (see Figure 4.1) or the sensing elements
from the LTCC ring, which were glued with a conductive silver/epoxy adhesive
(see Figure 5.20 in Weilguni, 2013). Once the glued electrical connection was
broken, artefacts interfered the measurements.
The results of the study showed that nger forces in clarinet playing are
very low compared to nger forces on other musical instruments. Further-
more, these forces seem not to be restricted to only vertical movements of the
ngers. Attaching force sensors directly at the nger tips of the player, would
additionally allow to measure nger forces of the same player performing on
dierent types of clarinets, but also allows to use the setup to capture nger
forces on other instruments, like saxophone or ute. On wind instruments,
were key pads close the tone holes (e.g., on ute Fabre et al. 2012), it would be
interesting to investigate in how far a leakage of the pads inuences the nger
forces.
For the technical exercise task of the study, we repeated a task from a pre-
vious saxophone study (Hofmann and Goebl, 2014). This allows to compare
temporal eects of clarinet performances with those on the saxophone. Par-
ticipants performed a melody which was designed to be played by dierent
eectors (ngers only, tongue only, tongue-nger actions), under three dier-
ent tempi. Overall, clarinetists showed a higher timing precision in comparison
to the saxophonists, especially with tongue-only actions. Nevertheless, similar
trends were found in the results of both studies. For the slow tempo con-
dition, combined tongue-nger actions achieved the highest timing precision.
Although clarinetists' tongue-only actions showed the highest timing preci-
sion in all three tempo conditions, we again observed the eect that for the
fast tempo condition, combined tongue-nger actions were close to the CV of
nger-only actions. This is yet another indicator, that ngers play a dominant
role in the overall timing of woodwind performances. Taking the nger force
measurements from the technical exercise into account, the study suggests that
professional players are able to adjust their nger forces to very light ngering
technique (Fmean = 0.54 N) to achieve better timing.

Chapter 5
Application of performance
measurements to physics based
sound synthesis
In this chapter physical modelling sound synthesis is used to illustrate the
dierence between tongue articulation and non-tongue articulation in single-
reed woodwind performance. The parameter changes observed in the sensor
reed measurements on saxophone and clarinet in Chapter 2.3 will be applied
to a lumped mass-spring model (Chatziioannou and van Walstijn, 2012), in
order to simulate the two dierent articulation techniques used by professional
players.
5.1 Introduction
Physical modelling of acoustic instruments has various applications. On one
hand, these models are used to explain the behaviour of real instruments
(Välimäki, 2004; Facchinetti et al., 2003; Dalmont et al., 2005), the inter-
action between player and instrument (Barthet et al., 2010; Van Walstijn and
Avanzini, 2007; Almeida et al., 2013) or help to analyse the inuence of mate-
rial properties to the radiated sound (Chatziioannou, 2010, see Chapter 6.6).
On the other hand, physical modelling is a promising approach to synthesize
sound for musical applications (Roads, 1996; Cook, 2002).
Based on acoustical measurements on single-reed instruments (e.g., Boutil-
93
Chapter 5. Application of performance measurements to physics based sound
94 synthesis
lon and Gibiat, 1996; Nederveen, 1998; Dalmont et al., 2003; Gazengel et al.,
2007), physical models for computer-based sound synthesis of steady-state
tones have been developed (Chatziioannou, 2010; Karkar et al., 2012; Van Wal-
stijn and Avanzini, 2007; Facchinetti et al., 2003).
Sounds, synthesized using a physical modelling technique can be shaped by
changing the control parameters of the model. This allows to modulate the
synthesized sound in a similar way as controlling a real instrument. Of partic-
ular interest is the control of the player over the instrument during expressive
performance. The way players shape note onsets and note osets during saxo-
phone and clarinet performance is essential for a high level of expressiveness
in their performance (see Chapter 2.1.1).
The embouchure of the player together with the reed-mouthpiece system
can be described as a non-linear excitation mechanism that is coupled to a
resonance tube (Chatziioannou and van Walstijn, 2012). During note onsets
and note osets in expressive woodwind performance, professional musicians
use various articulation techniques to control the oscillations of the instru-
ment. In eect, transient phenomena arise at the beginning or at the end of a
tone, or during note-to-note transitions. The underlying physics of these short
transients is hard to measure under real playing conditions.
In the last years the interest in modelling transient behaviour for single-reed
instruments grew (Sterling et al., 2009; Bergeot et al., 2012; Guillemain and
Vergez, 2012). Two dierent assumptions of the tonguing related parameter
changes were discussed earlier (Chapter 2.3, Hofmann et al. (2012a): Ducasse
(2003) described the damping eect of the tongue to the reed and its force
to change the equilibrium position of the reed. In contrast to that, Sterling
et al. (2009) modelled the tongue as a gate to the mouthpiece, which prevents
the air-stream to enter. The performance measurements presented in Chapter
2.3.3 and 3.2.2 showed that there is a clear distinction between the signals (reed
bending, mouthpiece pressure, blowing pressure) captured for both articulation
techniques. Following, the results of the performance observations will be used
to derive control parameters for tongued and air-separated tone transitions,
that can be applied to physical modelling sound synthesis.

synthesis 95
5.2 Physical modelling
In this experiment, the physical model of a clarinet, presented in Chatziioan-
nou and van Walstijn (2012), will be adopted to simulate tongued tone tran-
sitions and air-separated tone transitions for the saxophone and the clarinet.
This model consists of a lumped mass-spring oscillator, coupled to the impulse
response of an appropriate resonator for saxophone (simple conical tube) or
clarinet (straight cylindrical tube). The aim of this study is to model artic-
ulation by only modifying existing model parameters. No complexity will be
added to the model. In a rst step, new model parameters for steady-state
sounds were estimated through inverse modelling. Mouthpiece pressure mea-
surements of a steady portion of the sound (see Chapter 2.3.3, Figure 2.12)
were used for this process. Details about the procedure of estimating the model
parameters for saxophone and clarinet can be found in the related publications
1
(Chatziioannou and Hofmann, 2013, 2015) .
5.2.1 Modelling articulation on single-reed woodwind in-

struments
A rst propose on how to model tongue and non-tongue articulation on the
saxophone was made in Hofmann et al. (2012a). Additional measurements on
the clarinet (given in Chapter 2.3.3) and further simulations (Chatziioannou
and Hofmann, 2013, 2015) were required to constitute these ndings. The
following section summarizes the parameter settings that were extracted from
these measurements and simulations and shows how they can be applied to
physics based sound synthesis.
Tongue articulation
To model a portato tongue stroke to the vibrating reed, a tongue-reed contact
duration of 30 ms is used (see Chapter 3.2.2, Characteristics of articulation
techniques ). In reference to the observed eects of the player's tongue to
1I am second author in both publications and contributed to the denition of the articu-
lation varied parameters and provided the processed data of the performance measurements.
Vasileios Chatziianou programmed the physical model and developed the mathematical de-
scription, which is not part of this thesis.
96 synthesis
1e+06
Damping (1/s)
1e+04
1e+02
0.20
Effective mass (kg)
0.10
0.00
Reed displacement (mm)
0.1
0.3
0.5
0.00 0.05 0.10 0.15

Time (s)
Figure 5.1: Tongue-reed contact is modelled by increasing the damping (top), the
eective mass (middle), and the reed displacement (bottom) of the mass-spring os-
cillator for 30 ms.
the vibrating sensor reed, described in Chapter 2.3.3 (Extracted parameters:
Tongued tone transition ), the following parameters of the physical model are
adjusted:
• increase of the internal damping of the mass-spring oscillator
• increase of the eective mass of the oscillator
• displacement of the equilibrium of the oscillator towards the mouthpiece
The blowing pressure remains approximately constant in this case, as veri-
ed by the measurements from Chapter 2.3.3 (Figure 2.13, left). The variation
of the related model parameters is shown in Figure 5.1.
Air-separated tones
The sensor reed measurements in Chapter 2.3.3 (Extracted parameters: Air-
separated tones ) showed that in the case of air-separated tone transitions, the
reed vibrations were not directly manipulated by the player's tongue. To model
air-separated tone transitions, only one parameter of the model is adjusted:

synthesis 97
4.0
3.5
3.0
2.5
2.0
0.0 0.1 0.2 0.3 0.4 0.5

Time (s)
Figure 5.2: Air-separated tones tones are modelled by a modulation of the blowing
pressure.
• reduction of blowing pressure
Figure 5.2 shows the variation of the blowing pressure parameter of the
model.
5.2.2 Results1
The simulated pressure and reed displacement signals are plotted in Figure 5.3
and Figure 5.4, showing both modelled articulation techniques.
The presented model parameters make a clear distinction between the two
articulations. The synthesized signals show similarities with the measurements
presented earlier in Chapter 2.3. In particular, in both Figures concerning the
clarinet, Figure 2.13 (measurement, page 27) and Figure 5.3 (model) (Fig-
ure 2.14 and 5.4 for the saxophone), the modulation of the reed oscillations
were softer when the tongue was used, because of the direct but short inter-
action with the reed. In the case of the model, the reed displacement was
calculated (Figure 5.3, top), whereas during the experiments the bending of
the reed was measured (Figure 2.13, top). These two signals are not directly
comparable. As a consequence, the measurements can only be used in a qual-
itative prediction of the motion of the reed.
Also for the mouthpiece pressure, a characteristic pressure envelope for each
articulation technique was visible in the measured signals. This àrticulation
1 as published in Hofmann et al. (2013a); Chatziioannou and Hofmann (2013, 2015)

98 synthesis
signature' is also present in the simulated pressure signals of Figure 5.3 (bot-
tom).
Finally, the increase in the reed damping caused by the contact with the
tongue results in a smoother synthesised pressure signal during the transient,
in comparison to the case of air-separated tones. Such an eect has also been
observed in the experimental measurements (see Figure 5.1). Sound exam-
ples generated by the model are online available at: http://iwk.mdw.ac.at/

?page_id=148&sprache=2.
5.3 Discussion
Sensor-based measurements during professional woodwind performance are a
useful tool to collect parameter variations to control physical models. Under-
standing the player-instrument interactions helps to synthesize more realistic
sounds. Capturing player-instrument interactions during performance always
involves limitations in the amount of parameters to track. It is a matter of
nding the right balance between adding sensor technology to the acoustic
instrument and not distracting the player. Therefore observations are limited
to a specic parameter set.
From the obtained blowing pressure, reed bending and mouthpiece pres-
sure measurements (Chapter 2.3), useful variations of articulatory controlled
parameters for the model were derived. An application of these parameter
changes to the model allowed to reproduce the measured pressure signals, at
least qualitatively, for steady-state and during transient oscillations (Hofmann
et al., 2012a; Chatziioannou and Hofmann, 2013).
In future work this approach will be enhanced towards the simulation of nu-
merous articulation techniques. To develop a more robust re-synthesis routine,
the reed bending signal needs to be directly correlated with reed displacement.
Furthermore, to ensure the quality of the re-synthesized sounds, listening tests
may be used as a measurement tool. These tests can range from sound com-
parisons between real recordings and synthetic sounds, to discrimination tasks
within the pool of synthesized sounds. If participants are able to discriminate
synthesized articulation techniques with the same precision as with recorded
sound samples, this would verify the variation of the model parameters.
synthesis 99
4e−04
4e−04
Reed displacement (m)

2e−04
2e−04
0e+00
0e+00

4
4
2
2
0
0
−2
−2
−4
−4
0.00 0.05 0.10 0.15 0.00 0.05 0.10 0.15
Time (s) Time (s)
Figure 5.3: Simulated reed displacement (top) and resulting mouthpiece pressure (bot-
tom) for a single note transition on the clarinet. Modelling results of tongue separated
tones (left) and air-separated tones (right).
2e−04
2e−04
0e+00
0e+00

1.5
1.5
0.5
0.5
−0.5
−0.5
−1.5
−1.5
0.10 0.15 0.20 0.10 0.15 0.20 0.25

Time (s) Time (s)
Figure 5.4: Simulated reed displacement (top) and resulting mouthpiece pressure (bot-
tom) for a single note transition on the saxophone. Modelling results of tongue sep-
arated tones (left) and air-separated tones (right).
100 synthesis
Chapter 6
Conclusion and future work
The research presented in this thesis investigated the interaction between mu-
sician and acoustic instrument in saxophone and clarinet performance. The
specic focus was tongue articulation measurements and nger force measure-
ments on the saxophone and the clarinet.
Within this dissertation, a unique method for capturing reed oscillations
was developed. The main objective was to monitor tongue-reed interaction
during performance. Reed modications, sensor calibration, and preliminary
tests are presented in Chapter 2. Although the sensor reeds were primarily de-
signed to track tongue articulation during empirical performance studies with
saxophone and clarinet, they also created precise measurements of articulation
parameters, showing the eects of the tongue to the reed.
Three empirical performance studies (two production experiments and one
perception experiment) make up the majority of this thesis (Chapter 3 and
4). The corpus of data within the production experiments contains more than
40 hours of saxophone and clarinet performances. Sensor reeds were used
to monitor tongue articulation. To measure nger forces in clarinet playing,
Weilguni (2013) developed special ring-shaped force sensors for a Viennese
Clarinet at the Vienna Technical University.
As a nal step, the measured physical parameters of professional players'
articulation techniques were applied to a physical model in an attempt to
simulate transient behaviour for tone transitions (Chapter 5).
101
102 Chapter 6. Conclusion and future work
6.1 Summary of contributions
The main contributions of this thesis are:
• A measurement setup to capture reed oscillations during performance
was developed. After calibrating the reed sensor, it was possible to ex-
tract tonguing parameters from professional players. The observations
support a tonguing model proposed by Ducasse (2003), in which a tongue
stroke shows a damping eect to the vibrating reed, and its force changes
the equilibrium position of the reed during a tongued tone transition.
• Through simultaneous measurements of blowing pressure, mouthpiece
pressure, and reed bending, physical control parameters for tongued and
non-tongued articulation were gained. These physical control parame-
ters were applied to a physical model (Chatziioannou, 2010) to simulate
transient behaviour in tone transitions. The reproduced pressure signals
of the model showed a clear distinction between the two articulation
techniques.
• Considering that saxophonists have to coordinate tongue and nger ac-
tions during expressive performance, an empirical study (N = 19) to
examine temporal eects of dierent eector combinations was carried
out. Through the sensor reed measurements, it was possible to distin-
guish between tongued and non-tongued tones in the recorded sequences.
For slow tempi, the performance timing beneted from combined tongue-
nger actions. This supports the multiple-timer model from Ivry et al.
(2002), which predicts stabilized timing for combined eectors. However,
in the fast tempo condition, nger actions showed a dominant eect on
the overall timing (Hofmann and Goebl, 2014).
• A reproduction of the saxophone experiment with clarinetists (N = 23)
conrmed this observation: In the fast tempo condition, nger actions
play a dominant role on the timing. Even tough in the case of the clar-
inetists, tongue-only timing was superior than that of the saxophonists,
combined tongue-nger actions followed the timing of the ngers.
• In a listening study (N = 31), it was tested if expertise in saxophone

Chapter 6. Conclusion and future work 103
playing helps to facilitate the discrimination of articulation techniques.
Participants with a dierent background in music making attended the
experiment (saxophonists, musicians not playing a wind instrument, non-
musicians). Independent of the listener's group, errors occurred when
legato articulation and portato articulation had to be discriminated, nev-
ertheless saxophonists showed the best results for this task. In reference
to the motor theory of speech perception (Galantucci et al., 2006), the
results of the experiment indicate that the link between production and
perception of speech may also apply to the production and perception of
articulation in saxophone music (Hofmann and Goebl, 2014).
• To examine nger forces with clarinet students (N = 17) and profes-
sional clarinetists (N = 6), an empirical study comprised of a variety of
performance conditions was designed and conducted. In comparison to
nger forces reported for other instruments (Kinoshita and Obata, 2009;
Hori et al., 2013; Parlitz et al., 1998), the measured average nger forces
for expressive playing was low (Fmean = 1.21 N). Finger forces measured
during a technical exercise task was even lower (Fmean = 0.64 N). The
group of professional clarinetists showed very light nger forces to achieve
precise timing in the technical exercise (Fmean = 0.54 N). These observa-
tions indicate that professional players use very light ngering technique
if possible.
6.2 Future work
The sensor instruments developed and used in this research enabled an investi-
gation of ne motor control of clarinet and saxophone players. There are some
technical aspects with the sensor technology that can be optimised in future
research.
During the nger force measurements on saxophone and clarinet, it turned
out that both setups showed weaknesses. On the saxophone, the standard
industry foil sensors attached to the pearl buttons were not touched by all
players. Here, it might be a possibility to place the sensor inside the key
between the pad and the cup. The ring shaped force sensors on the clarinet
showed two weaknesses. First, the polymer rings glued on three glass balls fell
o. Second, the electric conductive silver/epoxy adhesive connection between
the sensing elements and the Low Temperature Co-red Ceramic (LTCC) ring
broke in some cases.
The developed sensor reed has only one strain gauge sensor attached. This
limits the observations to only one specic area of the reed. For example, this
does not allow for the observation of torsional modes that may occur in the
vibration patterns of clarinet reeds, as reported in Pinard et al. (2003). Using
a high-speed camera and an articial blowing machine for further calibration
would help to monitor the tip of the sensor reed. The position and the amount
of strain gauge sensors on the reed could then be optimized. Furthermore, it
provides details about the link between reed bending measurements and the
tip opening.
Compared to more traditional synthesis methods that attempt to repli-
cate sound spectra, physical modelling aims to simulate the sound generation
mechanism of musical instruments. This oers a more realistic reproduction
of waveforms and also allows to include player actions that are controlling the
involved physical parameters. With the link between reed bending and tip
opening, it would be possible to directly compare the behaviour of a physical
model with the measured sensor reed signal, captured in a human performance.
This could be a way to further optimize the physical model towards realistic
tone transitions, which are capable of simulating a variety of articulation tech-
niques, similar to the repertoire of professional players.
Another way to verify sounds of a model, apart from comparing physical
parameters, could be listening tests. Similar, to the AB-X test presented in
Chapter 3.3, the modelled sounds could be in a pool of stimuli. If participants
can discriminate modelled articulation techniques with the same precision as
recorded articulation techniques, this would additionally verify the model pa-
rameters applied for the tone transitions.
The combination of measurement techniques from music acoustics and the
empirical approach from performance science research would allow a compari-
son of style between jazz and classical soloists. Applying stylistically grouped
control parameters to a physical model, may be the key to re-synthesize more
expressive tone transitions in comparison to traditional sample playback.

Chapter 6. Conclusion and future work 105
Apart from looking only at physical parameter changes in solo performances,
such sensor instruments can be used to investigate a large variety of dierent
performance situations. A future application may be to investigate the role
of each musician in a music ensemble. Especially for the saxophone, which is
primarily played in jazz music, this might be an approach to better understand
groove in jazz ensembles (Prögler, 1995). A combination of sensor instruments,
video technology, and acceleration sensors, also on the musician's body, gives
the opportunity to study ensemble synchronisation at various levels, ranging
from large pick-up gestures to the ne motor control of each individual player.
Bibliography
Almeida, A., Chow, R., Smith, J., and Wolfe, J. ( 2009). The kinetics and
acoustics of ngering and note transitions on the ute, The Journal of the
Acoustical Society of America 126, 15211529.
Almeida, A., George, D., Smith, J., and Wolfe, J. ( 2013). The clarinet: How
blowing pressure, lip force, lip position and reed hardness aect pitch,
sound level, and spectrum, The Journal of the Acoustical Society of America
134, 22472255.
Almeida, A., Lemare, J., Sheahan, M., Judge, J., Auvray, R., Dang, K., John,
2010). Clarinet parame-

S., Georoy, J., Katupitiya, J., Santus, P., et al. (
ter cartography: automatic mapping of the sound produced as a function of
blowing pressure and reed force, in Proceedings of the International Sym-
posium on Music Acoustics (Sydney, Australia).
Altenmüller, E. and Jabusch, H.-C. ( 2004). Chronische Schmerzen beim Mu-

sizieren, Das Orchester 7/8 1722.
Baader, A., Kazennikov, O., and Wiesendanger, M. ( 2005). Coordination
of bowing and ngering in violin playing, Cognitive Brain Research 23,

436443.
Barthet, M., Guillemain, P., Kronland-Martinet, R., and Ystad, S. ( 2010).

From clarinet control to timbre perception, Acta Acustica united with
Acustica 96, 678689.
Bates, P. ( 1984). The New Grove Dictionary of Musical Instruments, chapter

Tonguing, edited by Sadie, S. (London: Macmillan Press Ltd.).
107
108 BIBLIOGRAPHY
1983). Analysis and synthesis of musical

Bengtsson, I. and Gabrielsson, A. (
rhythm, Studies of music performance 39, 2760.
Bergeot, B., Almeida, A., Vergez, C., and Gazengel, B. ( 2012). Attack tran-
sients in a clarinet model with time-varying blowing pressure, in Proceedings
of the Acoustics 2012 Nantes Conference, 19811986 (Nantes, France).
Bertsch, M. and Mayer, A. ( 2005). A new 3d transducer for measuring the

trumpet mouthpiece force, in Proceedings of the Forum Acusticum, 2330
(Budapest, Hungary).
2009). ...in Liebe zerossenes Gefühl Die Klarinette, Symposium

Birsak, K. (
im Rahmen der 30. Tage Alter Musik in Herne 2005, chapter 'Zur Geschichte
des vollen Klarinettentones', 4263 (München: Musikverlag Katzbichler).
Bishop, L. and Goebl, W. ( 2014). Context-specic eects of musical expertise

on audiovisual integration, Frontiers in Psychology: Cognitive Science 5,
114.
Boutillon, X. and Gibiat, V. ( 1996). Evaluation of the acoustical stiness
of saxophone reeds under playing conditions by using the reactive power
approach, The Journal of the Acoustical Society of America 100, 1178
1189.
2006). The aubio library at mirex 2006, Music Information

Brossier, P. M. (
Retrieval Evaluation eXchange 2006 13.
Caramiaux, B., Wanderley, M., and Bevilacqua, F. ( 2012). Segmenting and

parsing instrumentalists' gestures, Journal of New Music Research 41, 13
29.
Chatziioannou, V. ( 2010). Forward and inverse modelling of single-reed
woodwind instruments with application to digital sound synthesis, Ph.D.
thesis, Queen's University Belfast.
Chatziioannou, V. and Hofmann, A. ( 2013). Modeling articulation techniques

in single-reed woodwind instruments, in Proceedings of Meetings on Acous-
tics, volume 19, 035059 (ASA, Montreal, Canada).

BIBLIOGRAPHY 109
Chatziioannou, V. and Hofmann, A. ( 2015). Physics-based analysis of artic-

ulatory player actions in single-reed woodwind instruments, Acta Acustica
united with Acustica 101, 292299.

Chatziioannou, V. and van Walstijn, M. ( 2012). Estimation of clarinet reed
parameters by inverse modelling, Acta Acustica united with Acustica 98,
629639.
Chen, J., Smith, J., and Wolfe, J. ( 2011). Saxophonists tune vocal tract reso-
nances in advanced performance techniques, The Journal of the Acoustical
Society of America 129, 415426.

Cook, P. R. ( 2002). Real Sound Synthesis for Interactive Applications (Welles-
ley: AK Peters, Ltd.).
Dahl, S. (2004). Playing the accent - comparing striking velocity and timing
in an ostinato rhythm performed by four drummers, Acta Acustica united
with Acustica 90, 762776.

Dahl, S., Bevilacqua, F., Bresin, R., Clayton, M., Leante, L., Poggi, I.,
and Rasamimanana, N. ( 2009). Musical Gestures: Sound, Movement, and
Meaning, chapter Gestures in performance, 3668 (New York: Routledge).
Dalmont, J., Gilbert, J., and Ollivier, S. ( 2003). Nonlinear characteristics of

single-reed instruments: Quasistatic volume ow and reed opening measure-
ments, The Journal of the Acoustical Society of America 114, 22532262.

Dalmont, J.-P., Gilbert, J., Kergomard, J., and Ollivier, S. ( 2005). An ana-
lytical prediction of the oscillation and extinction thresholds of a clarinet,
The Journal of the Acoustical Society of America 118, 32943305.

D'Ausilio, A., Pulvermüller, F., Salmas, P., Bufalari, I., Begliomini, C., and
Fadiga, L. ( 2009). The motor somatotopy of speech perception, Current

Biology 19, 381385.
Desmet, F., Nijs, L., Demey, M., Lesare, M., Martens, J.-P., and Leman,
M. ( 2012). Assessing a clarinet player's performer gestures in relation to

locally intended musical targets, Journal of New Music Research 41, 3148.
110 BIBLIOGRAPHY
Diethelm, U. 2011).
( Schmerzen beim Musizieren, Schweizerische
Gesellschaft für Musik-Medizin 5, 60.

Ducasse, E. (2003). A physical model of a single-reed wind instrument, in-
cluding actions of the player, Computer Music Journal 27, 5970.
Fabre, B., Auvray, R., Curtit, M., and Roosen, J.-Y. ( 2012). Measuring leak-
ages in ute pads, in Proceedings of the Acoustics 2012 Nantes Conference
(Nantes, France).
Facchinetti, M., Boutillon, X., and Constantinescu, A. ( 2003). Numerical and

experimental modal analysis of the reed and pipe of a clarinet, The Journal
of the Acoustical Society of America 113, 28742883.

Fadiga, L., Craighero, L., Buccino, G., and Rizzolatti, G. ( 2002). Speech
listening specically modulates the excitability of tongue muscles: a tms
study, European Journal of Neuroscience 15, 399402.

1997). Neurobiology of speech per-
Fitch, R. H., Miller, S., and Tallal, P. (
ception, Annual Review of Neuroscience 20, 331353.
Fletcher, N. H. ( 1979). Air ow and sound generation in musical wind in-
struments, Annual Review of Fluid Mechanics 11, 123146.

Forssberg, H., Eliasson, A., Kinoshita, H., Johansson, R., and Westling, G.
(1991). Development of human precision grip I: Basic coordination of
force, Experimental Brain Research 85, 451457.

2003). Music performance research at the millennium, Psy-
Gabrielsson, A. (
chology of music 31, 221272.
2006). The motor theory of

Galantucci, B., Fowler, C. A., and Turvey, M. T. (
speech perception reviewed, Psychonomic Bulletin & Review 13, 361377.
Gardner, J. (2010). Ultrasonic investigation of clarinet multiple articulation,

Ph.D. thesis, Arizona State University.
Gazengel, B., Guimezanes, T., Dalmont, J.-P., Doc, J., Fagart, S., and Léveillé,
Y. (2007). Experimental investigation of the inuence of the mechanical

BIBLIOGRAPHY 111
characteristics of the lip on the vibrations of the single reed, in Proceedings
of the International Symposium on Musical Acoustics (Barcelona, Spain).
Gingras, B. and McAdams, S. ( 2011). Improved score-performance match-
ing using both structural and temporal information from midi recordings,
Journal of New Music Research 40, 4357.

Goebl, W. and Palmer, C. (2008). Tactile feedback and timing accuracy in
piano performance, Experimental Brain Research 186, 471479.
Goebl, W. and Palmer, C. ( 2009). Finger motion in piano performance:
Touch and tempo, in International Symposium on Performance Science
(ISPS), 6570 (Auckland, New Zealand).
Goebl, W. and Palmer, C. ( 2013). Temporal control and hand movement
eciency in skilled music performance, PLoS ONE 8, e50901.

Goolsby, T. ( 1997). Verbal instruction in instrumental rehearsals: A com-
parison of three career levels and preservice teachers, Journal of Research
in Music Education 45, 2140.

Grachten, M., Gasser, M., Arzt, A., and Widmer, G. ( 2013). Automatic align-
ment of music performances with structural dierences., in Proceedings of
the 14th International Society for Music Information Retrieval Conference,
607612 (Curitiba, Brazil).
Grosshauser, T. and Troester, G. ( 2013). Sensor based hand weight and
pressure measurements in trombone playing, in Proceedings of the Stock-
holm Music Acoustics Conference 2013, SMAC 13, edited by R. Bresin and
A. Askenfeldt, 398402 (KTH Royal Institute of Technology, Stockholm,
Sweden).
Guillemain, P. and Vergez, C. ( 2012). Decay transients in single reed wind

instruments, in Proceedings of the Acoustics 2012 Nantes Conference, 1975
1979 (Nantes, France).
Guillemain, P., Vergez, C., Ferrand, D., and Farcy, A. ( 2010). An instru-
mented saxophone mouthpiece and its use to understand how an experienced
musician plays, Acta Acustica united with Acustica 96, 622634.

112 BIBLIOGRAPHY
Hall, D. ( 1991). Musical acoustics, Physics Series (California: Brooks/Cole
Publishing).
Haueisen, J. and Knösche, T. R. ( 2001). Involuntary motor activity in pi-
anists evoked by music perception, Journal of Cognitive Neuroscience 13,

786792.
Hofmann, A., Chatziioannou, V., Kausel, W., Goebl, W., Weilguni, M., and
Smetana, W. ( 2012a). The inuence of tonguing on tone production with

single-reed instruments, in Proceedings of the 5th Congress of the Alps Adria
Acoustics Association (Zadar, Croatia).
Hofmann, A., Chatziioannou, V., Weilguni, M., Goebl, W., and Kausel, W.
(2013a). Measurement setup for articulatory transient dierences in wood-

wind performance, in Proceedings of Meetings on Acoustics, volume 19,
035060 (Montreal, Canada).
Hofmann, A. and Goebl, W. ( 2014). Production and perception of legato,
portato and staccato articulation in saxophone playing, Frontiers in Psy-
chology: Cognitive Science 5, 110.

Hofmann, A., Goebl, W., and Weilguni, M. ( 2013b). Evaluating a wavelet-
based analysis of sensor reed signals for performance research, in Proceedings
of the Stockholm Music Acoustics Conference 2013, SMAC 13, edited by
R. Bresin and A. Askenfeldt, 398402 (KTH Royal Institute of Technology,
Stockholm, Sweden).
Hofmann, A., Goebl, W., Weilguni, M., Mayer, A., and Smetana, W. ( 2012b).
Measuring tongue and nger coordination in saxophone performance, in
12th International Conference on Music Perception and Cognition (ICMPC)
and 8th Triennial Conference of the European Society for the Cognitive Sci-
ences of Music (ESCOM), 442445 (Thessaloniki, Greece).
Hofmann, A., Goebl, W., Weilguni, M., and Smetana, W. ( 2013c). Zooming
into saxophone performance: Tongue and nger coordination, in Proceed-
ings of the International Symposium on Performance Science, edited by
A. Williamon and W. Goebl, 289294 (Brussels, Belgium).

BIBLIOGRAPHY 113
Hori, M., Hosono, M., Takahashi, H., Matsumoto, K., and Shimoyama, I.
(2013). 3-axis ngertip force during playing the string instrument, in
Proceedings of 2013 Transducers and Eurosensors XXVII: The 17th Inter-
national Conference on Solid-State Sensors, Actuators, and Microsystems,
27452748 (IEEE, Barcelona, Spain).
Ivry, R., Richardson, T., and Helmuth, L. ( 2002). Improved temporal sta-
bility in multieector movements., Journal of Experimental Psychology:
Human Perception and Performance 28, 7292.

Jaurová, D. ( 2013). Artikulation am Saxophon, (unpublished diploma the-
sis), Universität für Musik und darstellende Kunst, Wien.
Jensenius, A., Wanderley, M., Godøy, R., and Leman, M. ( 2010). Musical Ges-
tures: Sound, Movement, and Meaning, chapter Musical gestures: concepts
and methods in research, 1235 (New York: Routledge).
Karkar, S., Vergez, C., and Cochelin, B. ( 2012). Oscillation threshold of
a clarinet model: a numerical continuation approach, The Journal of the
Acoustical Society of America 131, 698707.

Kergomard, J. ( 1995). Mechanics of Musical Instruments, chapter Elemen-
tary considerations on reed-instrument oscillations, 229290 (New York:
Springer).
Kinoshita, H. and Obata, S. ( 2009). Left hand nger force in violin play-
ing: Tempo, loudness, and nger dierences, The Journal of the Acoustical
Society of America 126, 388395.

Kinoshita, H.and Obata, S., Hirano, T., Ohsawa, C., and Ito, T. ( 2012).
Expert-novice dierence in string clamping force in violin playing, in 12th
International Conference on Music Perception and Cognition (ICMPC) and
8th Triennial Conference of the European Society for the Cognitive Sciences
of Music (ESCOM), 532533 (Thessaloniki, Greece).
Koch, E. ( 1989). Schule für Klarinette in B: Deutsches und französisches Sys-

tem (Leipzig: Deutscher Verlag für Musik).
114 BIBLIOGRAPHY
Krautgartner, K. ( 1982). Untersuchungen zur Artikulation bei Klarinettenin-

strumenten im Jazz, Ph.D. thesis, Universität zu Köln.
Kronland-Martinet, R., Morlet, J., and Grossmann, A. ( 1987). Analysis of

sound patterns through wavelet transforms, International Journal of Pat-
tern Recognition and Articial Intelligence 1, 273302.
Legere, G. ( 2000). Oriented polymer reeds for musical instruments, US
Patent 6,087,571.
Legere, G., Kortschot, M., et al. ( 2011). Oriented polymer reeds for woodwind
instruments, US Patent 7,902,443.
Leman, M. and Godøy, R. ( 2010). Musical gestures: Sound, movement, and
meaning, chapter 'Why Study Musical Gestures?', 311 (New York: Rout-
ledge).
Liberman, A. M. and Mattingly, I. G. ( 1985). The motor theory of speech

perception revised, Cognition 21, 136.
Liebman, D. ( 1989). Developing a personal saxophone sound (Medeld: Dorn
Publications).
Manto, M., Bower, J. M., Conforto, A. B., Delgado-García, J. M., da Guarda,
S. N. F., Gerwig, M., Habas, C., Hagura, N., Ivry, R. B., Mariën, P., et al.
(2012). Consensus paper: Roles of the cerebellum in motor control the
diversity of ideas on cerebellar involvement in movement, The Cerebellum
11, 457487.
Mauz, R. ( 2011). School of Velocity (Schule der Geläugkeit) (Mainz: Schott
Music).
Michaels, J. ( 1999). Methodische Schule der klarinettistischen Gritechnik
(Frankfurt: Zimmermann Verlag).
Nederveen, C. ( 1998). Acoustical Aspects of Woodwind Instruments (DeKalb:
Northern Illinois University Press).

BIBLIOGRAPHY 115
Nemoto, K. and Arino, H. ( 2007). Hand and upper extremity problems in
wind instrument players in military bands, Medical Problems of Performing
Artists 22, 6769.

Palmer, C., Carter, C., Koopmans, E., and Loehr, J. ( 2007). Movement,
planning, and music: Motion coordinates of skilled performance, in Pro-
ceedings of the International Conference on Music Communication Science,
119122 (Sydney, Australia).
Palmer, C., Koopmans, E., Carter, C., Loehr, J., and Wanderley, M. ( 2009a).
Synchronization of motion and timing in clarinet performance, in Interna-
tional Symposium on Performance Science (Auckland, New Zealand).
Palmer, C., Koopmans, E., Loehr, J., and Carter, C. ( 2009b). Movement-
related feedback and temporal accuracy in clarinet performance, Music
Perception 26, 439449.

Paradzinets, A., Harb, H., and Chen, L. ( 2006). Use of continuous wavelet-
like transform in automated music transcription, in Proceedings of the Eu-
ropean Signal Processing Conference (Florence, Italy).
Parlitz, D., Peschel, T., and Altenmüller, E. ( 1998). Assessment of dynamic

nger forces in pianists: eects of training and expertise, Journal of Biome-
chanics 31, 10631067.

Pay, A. ( 1995). The Cambridge Companion to the Clarinet, chapter The me-
chanics of playing the clarinet, 107122 (Cambridge: Cambridge University
Press).
Percival, D. and Walden, A. ( 2006). Wavelet Methods for Time Series Anal-
ysis, Cambridge Series In Statistical And Probabilistic Mathematics (Cam-
bridge: Cambridge University Press).
Pinard, F., Laine, B., and Vach, H. ( 2003). Musical quality assessment of
clarinet reeds using optical holography, The Journal of the Acoustical So-
ciety of America 113, 17361742.

Prögler, J. A. ( 1995). Searching for swing: Participatory discrepancies in the
jazz rhythm section, Ethnomusicology 2154.

116 BIBLIOGRAPHY
R Core Team ( 2012). R: A Language and Environment for Statistical Com-
puting, R Foundation for Statistical Computing, Vienna, Austria.
Rabiner, L. and Juang, B. ( 1993). Fundamentals of Speech Recognition,
Prentice-Hall Signal Processing Series: Advanced monographs (Englewood
Clis: Prentice Hall).
Rasamimanana, N. ( 2012). Towards a conceptual framework for exploring
and modelling expressive musical gestures, Journal of New Music Research
41, 312.
Repp, B. H. ( 2005). Sensorimotor synchronization: a review of the tapping
literature, Psychonomic Bulletin & Review 12, 969992.
Roads, C. ( 1996). The computer music tutorial (Massachusetts: MIT Press).
Scavone, G., Lefebvre, A., and da Silva, A. ( 2008). Measurement of vocal-

tract inuence during saxophone performance, The Journal of the Acousti-
cal Society of America 123, 23912400.

Schoonderwaldt, E. and Demoucron, M. ( 2009). Extraction of bowing pa-
rameters from violin performance combining motion capture and sensors,
The Journal of the Acoustical Society of America 126, 26952708.

Shackleton, N. ( 1995). The Cambridge Companion to the Clarinet, chapter
The Development of the Clarinet, 1632 (Cambridge: Cambridge University
Press).
Spahn, C., Richter, B., and Altenmüller, E. ( 2011). MusikerMedizin: Di-
agnostik, Therapie und Prävention von musikerspezischen Erkrankungen
(Stuttgart: Schattauer).
Sterling, M., Dong, X., and Bocko, M. ( 2009). Pitch bends and tonguing ar-
ticulation in clarinet physical modeling synthesis, in Proceedings of the 2009
IEEE International Conference on Acoustics, Speech and Signal Processing,
289292 (IEEE Computer Society, Taipei, Taiwan).
Tubiana, R. et al. ( 2005). Functional anatomy of the hand / Movements of

the ngers / Fundamental positions for instrumental musicians (anniversary
article), Medical Problems of Performing Artists 20, 183.

BIBLIOGRAPHY 117
Tzanetakis, G., Essl, G., and Cook, P. ( 2001). Audio analysis using the
discrete wavelet transform, in Proceedings of the Conference in Acoustics
and Music Theory Applications, 318323 (Skiathos, Greece).
Välimäki, V. (2004). Physics-based modeling of musical instruments, Acta

Acustica united with Acustica 90, 611617.
Van Walstijn, M. and Avanzini, F. ( 2007). Modelling the mechanical response

of the reed-mouthpiece-lip system of a clarinet. Part II: A lumped model
approximation, Acta Acustica united with Acustica 93, 435446.

van Walstijn, M. and de Sanctis, G. ( 2014). Adaptive calibration of a three-
microphone system for acoustic waveguide characterization under time-
varying conditions, The Journal of the Acoustical Society of America 135,

917927.
Wanderley, M., Vines, B. W., Middleton, N., Mckay, C., and Hatch, W.
(2005). The musical signicance of clarinetists' ancillary gestures: An ex-
ploration of the eld, Journal of New Music Research 34, 97113.
Wehle, R. ( 2007). Clarinet Fundamentals Volume 2: Systematic Fingering
Course, Clarinet Fundamentals (Mainz: Schott Music).
Weilguni, M. ( 2013). Force sensors for the measurement of nger forces in
clarinet playing, Ph.D. thesis, Vienna University of Technology.
Weilguni, M., Hofmann, A., Smetana, W., and Goebl, W. ( 2013). Fabri-
cation and mechanical characterization of ring-shaped LTCC/HTCC nger
force sensors used for performance studies in clarinet playing, in The 17th
International Conference on Solid-State Sensors, Actuators and Microsys-
tems (Barcelona, Spain).
Weilguni, M., Smetana, W., Hofmann, A., Radosavljevic, G., Goebl, W., and
Nicolics, J. ( 2012). A ring-shaped LTCC/HTCC sensor for detection of
nger forces in clarinet playing, in IEEE Sensors 2012 Conference (Taipei,
Taiwan).
Window, A. ( 1992). Strain Gauge Technology, An Objective publication (New

York: Springer).
118 BIBLIOGRAPHY
2007). When the brain plays

Zatorre, R. J., Chen, J. L., and Penhune, V. B. (
music: auditorymotor interactions in music perception and production,
Nature Reviews Neuroscience 8, 547558.

Appendix A
Additional material saxophone
experiments
119
120 Chapter A. Additional material saxophone experiments
Note-onsets (TRR) Note-osets (TRC)
Prec. Rec. F-Meas. Prec. Rec. F-Meas.
Portato Slow 0.9793 0.9861 0.9827 0.9862 0.9931 0.9896
Portato Med. 0.8767 0.8828 0.8797 0.9247 0.9310 0.9278
Portato Fast 0.9861 0.9726 0.9793 0.9653 0.9521 0.9586
Staccato Slow 0.9388 0.9517 0.9452 0.9456 0.9586 0.9521
Staccato Med. 0.9728 0.9795 0.9761 0.9864 0.9932 0.9898
Staccato Fast 1.0000 0.9384 0.9682 0.9854 0.9310 0.9574
Legato Slow 0.9390 0.9625 0.9506
Legato Med. 0.9792 0.9691 0.9741
Legato Fast 0.9452 0.6900 0.7977
Sum 0.958 0.931 0.944 0.965 0.960 0.963
Table A.1: F-measure, precision and recall for wavelet based onset and oset de-
tection in the sensor reed signal. Results for a ±25 ms evaluation window are given
and further discussed in Section 3.2.1.
Chapter A. Additional material saxophone experiments 121
IOI − Similarity Pitch − Similarity

a) b)
0 0 0 0 0 0 0 0 0 0 0.020.030.050.110.29 1 1 1 1 1 0.290.110.05 1 0.290.110.05 1 0.290.110.05 1
15
15
0 0 0 0 0 0 0 0 0 0.020.030.050.110.29 1 0.29 0.050.050.050.050.110.29 1 0.050.110.29 1 0.050.110.29 1 0.05
0 0 0 0 0 0 0 0 0.020.030.050.110.29 1 0.290.11 0.110.110.110.110.29 1 0.290.110.29 1 0.290.110.29 1 0.290.11
0 0 0 0 0 0 0 0.020.030.050.110.29 1 0.290.110.05 0.290.290.290.29 1 0.290.110.29 1 0.290.110.29 1 0.290.110.29
0 0 0 0 0 0 0.020.030.050.110.29 1 0.290.110.050.03 1 1 1 1 0.290.110.05 1 0.290.110.05 1 0.290.110.05 1
Performance Events
Performance Events
0 0 0 0 0 0.020.030.050.110.29 1 0.290.110.050.030.02 0.050.050.050.050.110.29 1 0.050.110.29 1 0.050.110.29 1 0.05
10
10
0 0 0 0 0.020.030.050.110.29 1 0.290.110.050.030.02 0 0.110.110.110.110.29 1 0.290.110.29 1 0.290.110.29 1 0.290.11
0 0 0 0.020.030.050.110.29 1 0.290.110.050.030.02 0 0 0.290.290.290.29 1 0.290.110.29 1 0.290.110.29 1 0.290.110.29
0 0 0.020.030.050.110.29 1 0.290.110.050.030.02 0 0 0 1 1 1 1 0.290.110.05 1 0.290.110.05 1 0.290.110.05 1
0 0.020.030.050.110.29 1 0.290.110.050.030.02 0 0 0 0 0.050.050.050.050.110.29 1 0.050.110.29 1 0.050.110.29 1 0.05
0.020.030.050.110.29 1 0.290.110.050.030.02 0 0 0 0 0 0.110.110.110.110.29 1 0.290.110.29 1 0.290.110.29 1 0.290.11
0.030.050.110.29 1 0.290.110.050.030.02 0 0 0 0 0 0 0.290.290.290.29 1 0.290.110.29 1 0.290.110.29 1 0.290.110.29
5
5
0.050.110.29 1 0.290.110.050.030.02 0 0 0 0 0 0 0 1 1 1 1 0.290.110.05 1 0.290.110.05 1 0.290.110.05 1
0.110.29 1 0.290.110.050.030.02 0 0 0 0 0 0 0 0 1 1 1 1 0.290.110.05 1 0.290.110.05 1 0.290.110.05 1
0.29 1 0.290.110.050.030.02 0 0 0 0 0 0 0 0 0 1 1 1 1 0.290.110.05 1 0.290.110.05 1 0.290.110.05 1
1 0.290.110.050.030.02 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0.290.110.05 1 0.290.110.05 1 0.290.110.05 1
5 10 15 5 10 15
Score Events Score Events
Final Cost Matrix

c) d)
1 1 1 1 1 1 1 1 1 1 1 0.970.980.990.98 0 ● ● ●
0
1 1
15
1 1 1 1 1 1 1 1 1 0.990.97 1 0.990.91 0 0.98

1 1 1 1 1 1 1 1 0.990.970.980.990.91 0 0.910.99
1 1 1 1 1 1 1 0.990.970.980.990.91 0 0.910.990.98 1 1 1
1 1 1 1 1 1 1 0.970.980.990.98 0 0.910.99 1 0.97
Performance Events
Reference index
1 1 1 1 1 0.990.97 1 0.990.91 0 0.980.990.980.97 1 −1
● ● ●
2 2 2
10
1 1 1 1 0.990.970.980.990.91 0 0.910.990.980.970.99 1
1 1 1 0.990.970.980.990.91 0 0.910.990.980.970.99 1 1
1 1 0.980.970.980.990.98 0 0.910.99 1 0.970.99 1 1 1 3 3 3
1 1 1 1 0.990.91 0 0.980.990.980.97 1 1 1 1 1
−2
0.98 1 0.990.990.91 0 0.910.990.980.970.99 1 1 1 1 1 ● ● ●
0.970.980.970.91 0 0.910.990.980.970.99 1 1 1 1 1 1
5
0.950.890.71 0 0.910.99 1 0.970.99 1 1 1 1 1 1 1

0.890.71 0 0.710.970.99 1 0.98 1 1 1 1 1 1 1 1
0.71 0 0.710.890.98 1 1 1 1 1 1 1 1 1 1 1
−3
0 0.710.890.950.970.98 1 1 1 1 1 1 1 1 1 1 ● ● ●
5 10 15 −3 −2 −1 0
Score Events Query index
Figure A.1: Pattern matching algorithm based on dynamic time warping. In this
example the performance data and score are identical. a) Similarity matrix based
on inter onset intervals (IOI) from recorded performance events and score events.
b) Pitch class distance between performance events and score events. c) Final Cost
Matrix is based on a one-matrix minus the IOI-Similarity Matrix multiplied with the
Pitch-Similarity Matrix. The algorithm searches the path with the lowest cost though
the matrix. The linear black line indicates a perfect match between performance and
score. d) The step pattern of the dynamic time warping algorithm denes rules during
the search. This Rabiner-Juang step pattern allows to omit up to three events.
122 Chapter A. Additional material saxophone experiments
IOI − Similarity Pitch − Similarity

a) b)
0 0 0 0 0 0 0 0.020.02 0 0.050.110.29 1 0.290.11 1 1 1 0.290.110.050.110.05 1 0.290.290.110.05 1 1 1
15
15
0 0 0 0 0 0 0.020.030.04 0 0.110.29 1 0.290.110.05 0.050.050.050.110.29 1 0.29 1 0.050.110.110.29 1 0.050.050.05
0 0 0 0 0 0 0.030.050.07 0 0.29 1 0.290.110.050.03 0.110.110.110.29 1 0.29 1 0.290.110.290.29 1 0.290.110.110.11
0 0 0 0 0 0 0.050.110.17 0 1 0.290.110.050.030.02 0.290.290.29 1 0.290.110.290.110.29 1 1 0.290.110.290.290.29
0 0 0 0 0 0.050.110.290.55 1 0.290.110.050.03 0 0 1 1 1 0.290.110.050.110.05 1 0.290.290.110.05 1 1 1
Performance Events
Performance Events
0 0 0 0 0.020.110.29 1 0.550.290.110.050.03 0 0 0 0.050.050.050.110.29 1 0.29 1 0.050.110.110.29 1 0.050.050.05
10
10
0 0 0 0.020.030.29 1 0.290.170.110.050.03 0 0 0 0 0.110.110.110.29 1 0.29 1 0.290.110.290.29 1 0.290.110.110.11
0 0 0.020.030.05 1 0 0 0 0 0 0 0 0 0 0 0.290.290.29 1 0.290.110.290.110.29 1 1 0.290.110.290.290.29
0 0 0.030.050.110.29 0 0 0 0 0 0 0 0 0 0 1 1 1 0.290.110.050.110.05 1 0.290.290.110.05 1 1 1
0 0.030.050.110.290.110.050.030.020.020.01 0 0 0 0 0 0.050.050.050.110.29 1 0.29 1 0.050.110.110.29 1 0.050.050.05
0.020.050.110.29 1 0.050.030.020.020.01 0 0 0 0 0 0 0.110.110.110.29 1 0.29 1 0.290.110.290.29 1 0.290.110.110.11
0.030.110.29 1 0.290.030.020.010.01 0 0 0 0 0 0 0 0.290.290.29 1 0.290.110.290.110.29 1 1 0.290.110.290.290.29
5
5
0.050.29 1 0.290.110.020.010.01 0 0 0 0 0 0 0 0 1 1 1 0.290.110.050.110.05 1 0.290.290.110.05 1 1 1
0.11 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 0.290.110.050.110.05 1 0.290.290.110.05 1 1 1
0.290.290.110.050.030.010.01 0 0 0 0 0 0 0 0 0 1 1 1 0.290.110.050.110.05 1 0.290.290.110.05 1 1 1
1 0.110.050.030.020.01 0 0 0 0 0 0 0 0 0 0 1 1 1 0.290.110.050.110.05 1 0.290.290.110.05 1 1 1
5 10 15 5 10 15
Score Events Score Events
Final Cost Matrix

c) d)
1 1 1 1 1 1 1 1 0.970.99 1 0.990.98 0 0.710.89 ● ● ●
0
1 1
15
1 1 1 1 1 1 1 0.97 1 0.99 1 0.91 0 0.980.99 1

1 1 1 1 1 1 1 0.980.990.95 1 0 0.910.990.99 1
1 1 1 1 1 1 0.980.990.910.45 0 0.910.990.980.99 1 1 1 1
1 1 1 1 1 1 0.990.98 0 0.840.910.99 1 0.97 1 1
Performance Events
Reference index
−1
1 1 1 1 0.990.970.91 0 0.980.980.990.980.97 1 1 1 ● ● ●
2 2 2
10
1 1 1 0.990.970.98 0 1 1 1 1 1 1 1 1 1
1 1 1 0.970.980.990.91 1 1 1 1 1 1 1 1 1
1 1 0.970.980.990.980.99 1 0.970.990.99 1 1 1 1 1 3 3 3
1 1 1 0.990.91 0 0.980.97 1 1 1 1 1 1 1 1
−2
0.98 1 0.990.91 0 0.910.970.99 1 1 1 1 1 1 1 1 ● ● ●
0.970.980.91 0 0.910.990.99 1 1 1 1 1 1 1 1 1
5
0.950.89 0 1 1 1 1 1 1 1 1 1 1 1 1 1
0.890.710.710.970.99 1 1 1 1 1 1 1 1 1 1 1
0.71 0 0.890.98 1 1 1 1 1 1 1 1 1 1 1 1
−3
0 0.890.950.970.980.99 1 1 1 1 1 1 1 1 1 1 ● ● ●
5 10 15 −3 −2 −1 0
Score Events Query index
Figure A.2: Pattern matching algorithm based on dynamic time warping. Example
with dierences between performance data and score. a) Similarity matrix based on
dierent inter onset intervals (IOI) from recorded performance events and expected
score events. b) Pitch class distances between performance events and score events.
c) Final Cost Matrix is based on a one-matrix minus the IOI-Similarity Matrix mul-
tiplied with the Pitch-Similarity Matrix. The black line shows the path found by the
algorithm to match the performance to the score. The algorithm searches the path
with the lowest cost trough the matrix. d) The step pattern of the dynamic time warp-
ing algorithm denes the rules during the search. This Rabiner-Juang step pattern
allows to omit up to three events.
Appendix B
Additional material clarinet
experiments
123
124 Chapter B. Additional material clarinet experiments

Tempo: 120 bpm
 44                    
                       

                       
8
        
Figure B.1: Melody designed for the warm-up task in the clarinet study (Chapter 4).
Slow Tempo Fast Tempo
Low Dynamics High Dynamics Low Dynamics High Dynamics
Low Register C A B H
High Register E F G D
Table B.1: Experimental design: 2 × 2 × 2 (tempo: slowfast, dynamics: piano

forte, register: lowhigh;), according score excerpts from Weber Clarinet Concerto,
can be found in Appendix B.2 (AD) and B.3 (EH).
Chapter B. Additional material clarinet experiments 125
 3 
Allegro

118
 4  
A
   
f  
 
  
Cadenza
 
159
  43 
B
     
  
Un poco ritenuto

170
C   43             
   
p
 
                           
Allegro Solo passinato

261
D   43    


       
         
3

ff
Figure B.2: Excerpts from the Clarinet Concerto No.1 in F minor (Op. 73) for
clarinet in Bb, from C.M.v.Weber, in order of appearance in the piece. Bar numbers
refer to the position in the rst movement.
                  

Adagio
      
1
E
    
p
        
      
 
 
            

     

     
Adagio
  
73
F
   
p 3 3
poco cresc. f
 2                                 
(Allegro)
         
48
G  4
p
sf p

                               
  
   
6
sf
             
269 (Allegro)
           
                      
H
   
             
      
Figure B.3: Excerpts from the Clarinet Concerto No.1 in F minor (Op. 73) for
clarinet in Bb, by C.M.v.Weber, in order of appearance in the piece. Bar numbers of
E) and F) refer to bar numbers in the second movement, bar numbers of G) and H)
refer the to third movement.
Tongue only Fingers + Tongue
 44                         
1 9 23
Figure B.4: Stimuli used for the technical exercise task in the clarinet study (Chap-
ter 4). 23-tone melody in B-at notation. Note numbers 18 require tonguing only.
Note numbers 923 require sequential key-depression by left-hand ngers.
Figure B.5: Two participants closing all 6 sensor equipped tone holes of the sensor
clarinet. Participant No. 2 (left) showed nger forces on an average level. In contrast
participant No. 4 (right) showed the highest peak nger forces measured during the
experiment (up to 12 N). Note the color of participant 4's nger tips. Figure B.6 and
B.7 show the related force measurement.
1.0
Reed Signal
−1.0 0.0
0 1 2 3 4 5
Ring Sensor 1 (N)
0.6
0.0
0 1 2 3 4 5
Ring Sensor 2 (N)
1.4
0.8
0.2
0 1 2 3 4 5
Ring Sensor 3 (N)
0.6
0.0
0 1 2 3 4 5
Ring Sensor 4 (N)
0.2 0.6 1.0
0 1 2 3 4 5
Ring Sensor 5 (N)
3.0
1.5
0.0
0 1 2 3 4 5
Ring Sensor 6 (N)
1.5
0.5
0 1 2 3 4 5
Time (s)
Figure B.6: Measured nger forces applied to the six ring shaped force sensors of the
clarinet. Participant 2 performs the excerpt A (bar 118) from the Weber Clarinet
Concerto No. 1 (nd score in Appendix Figure B.2).
Figure B.7: Measured nger forces applied to the six ring shaped force sensors of the
clarinet. Participant 4 performs the excerpt A (bar 118) from the Weber Clarinet
Concerto No. 1 (nd score in Appendix Figure B.2).
Figure B.8: Left-hand index nger force (in Newton) on clarinet, captured for par-
ticipants 18 performing the Weber excerpt A. The plot shows two force curves (N),
for the two captured trials under the highly expressive playing condition.
ticipants 916 performing the Weber excerpt A. The plot shows two force curves (N),
for the two captured trials under the highly expressive playing condition.
ticipants 1722 performing the Weber excerpt A. The plot shows two force curves
(N), for the two captured trials under the highly expressive playing condition.

Articulation and Finger Forces in Saxophone and Clarinet Playing (PDFDrive)

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Articulation and Finger Forces in Saxophone and Clarinet Playing (PDFDrive)

Uploaded by

Copyright:

Available Formats

Articulation and Finger Forces in

Saxophone and Clarinet playing

Am Institut für Wiener Klangstil der

Datum der Einreichung (submitted): 29. 1. 2015

duce virtuosic performances. While performing, they have to control various

articulation. This thesis puts a focus on tongue articulation measurements

and nger force measurements in saxophone and clarinet performance. Three

empirical studies as well as physical modelling sound synthesis address the

question of how players control the instrument to produce expressive tone

First, in a sound production experiment on the saxophone and the clarinet,

the participants had to perform melodies which require dierent techniques

to play the tones (tongue-only actions, nger-only actions, combined tongue

and nger actions) in three tempo conditions. A strain gauge sensor-equipped

fast tempo condition, the timing precision of combined tongue-nger actions

and clarinet performances.

Clarinet students and professional clarinetists performed two tasks (expressive

performance task, technical exercise task) on a sensor equipped Viennese Clar-

forces for the technical exercise task.

Third, in a listening experiment, it was questioned whether motor expertise

in music performance has an inuence on the ability to discriminate articu-

lation techniques in saxophone sounds. Listeners with dierent expertise in

1 published in Hofmann and Goebl (2014)

musicians) participated in an AB-X listening test. Their task was to discrim-

inate saxophone phrases containing legato, portato and staccato articulation.

All participants could easily discriminate between staccato articulation and

Finally, the articulation parameters obtained during the experiments were

applied to a physical model in an attempt to simulate tongued and non-tongued

1 published in Hofmann and Goebl (2014)

gen jahrelanges Training um saubere Tonübergänge und üssige Läufe im

ausdrucksstarken Spiel erzeugen zu können. Während des Spielens haben sie

Artikulationstechniken auch die Zungenbewegungen zu koordinieren. Saubere

Artikulation, ohne Nebengeräusche erfordert sehr viel Übung, da für einige

Artikulationstechniken die Zunge direkt das schwingende Blatt berührt. Der

Fokus dieser Arbeit liegt in der Untersuchung von Zungenartikulation und

Fingerkräften im Saxophon- und Klarinettenspiel.

In einer ersten empirischen Studie spielten Saxophonistinnen, Saxophon-

isten, Klarinettistinnen und Klarinetisten eine technische Übung, bei deren

Umsetzung verschiedene Spieltechniken koordiniert werden mussten (die Zun-

genbewegungen allein, die Fingerbewegungen allein, kombinierte Zungen- und

Fingerbewegungen). Es wurde ein spezielles Sensoreinfachrohrblatt entwickelt,

welches es erlaubt in den Blattschwingungen einen Zungen-Blattkontakt zu

erkennen. Die Ergebnisse konnten zeigen, dass in einem langsamen Spieltem-

po die Kopplung von Zungen- und Fingerbewegungen zu einer verbesserten

Rhythmik beitrugen. Allerdings war auch zu beobachten, dass im schnellen

Tempo die gekoppelten Bewegungen vorwiegend von den Fingerbewegungen

Eine zweite Studie mit studierenden und professionellen Klarinettenspielerin-

nen und Klarinettenspielern fokussierte auf die Fingerkräfte beim Schlieÿen

der Tonlöcher. Eine an der Technnischen Universität eigens dafür entwickelte

Wiener Klarinette mit Kraftsensoren wurde für diese Messungen verwendet.

Obwohl die einzelnen Fingerkraftprole beider Gruppen groÿe Überlappun-

gen zeigten, wurde dennoch deutlich, dass ausdrucksstarkes Spielen zu einem

erhöhten Einsatz von Fingerkräften führte. Beim Spielen einer technischen

Übung wurden demnach im Mittel geringere Fingerkräfte gemessen und ger-

Fall den geringsten Fingerkrafteinsatz.

1 veröentlicht in Hofmann und Goebl (2014)

zu einer verbesserten Unterscheidungsfähigkeit von verschieden artikulierten

Tonübergängen führt (legato, portato, staccato). Von den drei Teilnehmergrup-

pen (Laien, Musizierende die kein Saxophon spielen und Saxophonspielende)

zeigten die Saxophoninstinnen und Saxophonisten die besten Ergebnisse. Das

in Saxophonmusik möglicherweise ähnliche Verarbeitungsvorgänge im Gehirn

Abschlieÿend wurden die sich verändernden, physikalischen Parameter, wel-

den auf ein physikalisches Model übertragen. Durch diese Resynthetisierung

and nger force measurements in saxophone and clarinet performance. Three

the participants had to perform melodies which require dierent techniques

to play the tones (tongue-only actions, nger-only actions, combined tongue

and nger actions) in three tempo conditions. A strain gauge sensor-equipped

fast tempo condition, the timing precision of combined tongue-nger actions

in music performance has an inuence on the ability to discriminate articu-

lation techniques in saxophone sounds. Listeners with dierent expertise in

gen jahrelanges Training um saubere Tonübergänge und üssige Läufe im

Obwohl die einzelnen Fingerkraftprole beider Gruppen groÿe Überlappun-

1 veröentlicht in Hofmann und Goebl (2014)

der Klänge wurde verstärkt deutlich, welchen Einuss die Zungenartikulation

1 veröentlicht in Hofmann und Goebl (2014)

and analysis of nger forces in clarinet playing (P23248-N24). The FWF

encouraging words during lunch and coee breaks. Furthermore, I am very

for testing dierent sensor setups on the clarinet.

encouragement made me nally decide to study music. Prof. Michael Beil,

2.1.2 Instrument shapes and ngerings . . . . . . . . . . . . . 7

2.1.3 Tongue and nger coordination in woodwind performance 9

Empirical research in music performance has been identied as a growing eld

sound-modifying gestured and dierent torso and head movements as ancil-

inetists used larger nger movements and increased accelerations at nger-key