Professional Documents
Culture Documents
Articulation and Finger Forces in Saxophone and Clarinet Playing (PDFDrive)
Articulation and Finger Forces in Saxophone and Clarinet Playing (PDFDrive)
eingereichte Dissertation
von
Alex Michael Hofmann
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-
parameters at the same time, e.g., ngerings, blowing, lip position and tongue
transitions.
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
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
Second, nger forces applied to the tone holes of the clarinet were measured.
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
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 .
tone transitions.
unter anderem die Finger, die Atmung, den Lippendruck und für bestimmte
beeinusst waren. Somit kann man schlussfolgern, dass, obwohl die Zunge
den akustischen Tonbeginn steuert, die Finger einen stärkeren Einuss auf
1
die Spielpräzision haben .
ade die Gruppe der professionellen Spielerinnen und Spielern zeigte in diesem
lässt die Schlussfolgerung zu, dass bei der Unterscheidung von Artikulation
che eine Artikulation mit Zunge von der Artikulation ohne Zunge unterschei-
This research was carried out within a funded FWF project on Measurement
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
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
grateful to my supervisors Prof. Dr. Wilfried Kausel and Prof. Dr. Christoph
Reuter for supporting my research ideas and giving me helpful advise, espe-
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
would like to thank Alexander Mayer for his technical support in the labo-
ratory, Konstantin Zabranski for his support with signal processing, Laura
v
Furthermore, I could not have completed this thesis without the help of all
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
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
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
thank you for giving me the chance to develop my personal approach to jazz,
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!
Abstract i
Zusammenfassung iii
Acknowledgements v
Contents vii
1 Introduction 1
1.1 State of the art . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
wind performance . . . . . . . . . . . . . . . . . . . . . . 6
2.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.2 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.3.2 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
vii
2.3.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.3.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.4.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.4.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.2.1 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.3.1 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.2.1 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.2.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
ments . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.2.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
5.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6 Conclusion and future work 101
6.1 Summary of contributions . . . . . . . . . . . . . . . . . . . . . 102
Bibliography 107
A Additional material saxophone experiments 119
B Additional material clarinet experiments 123
Chapter 1
Introduction
in recent decades (Gabrielsson, 2003). There are multiple reasons that explain
musician turns a written musical score into pleasant sound events. On the
From the perspective of music acoustic research, the control that a player
tain player actions manipulate the resulting sound. The so called Musician-
acoustic instrument interaction has been named as a hot topic in recent acous-
goal directed body movements. These body movements have been categorised
1
2 Chapter 1. Introduction
et al. (2010) dened a gesture space for piano performance. He dened player
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
laboratory (Goebl and Palmer, 2008), Palmer et al. (2009b) showed that clar-
contact during dicult and fast musical tone sequences to achieve higher tem-
How much nger force do professional players apply to the tone holes? Is there
ger forces to the tone holes does not exist so far. In this research project an
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
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
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
clarinet and saxophone. The results of the experiments are supposed to solve
This thesis is structured into four main Chapters (25). Each Chapter has its
the single-reed, with nger actions at the tone holes simultaneously. Dierent
nger forces, key accelerations and reed signals are captured simultaneously.
1
Chapter 3 investigates the production and perception of dierent articula-
motor actions to perform the tones. Timing properties are investigated and
sounds is discussed.
In Chapter 4, the sensor clarinet which was developed at the Vienna Uni-
mance task, excerpts from Clarinet Concerto No.1 in F minor (Op. 73) for
jacent technical exercise task, the production experiment from the saxophone
study was repeated. Finger force proles, timing, and articulation techniques
enhance the model towards the simulation of tongued and air-separated tone
The research presented in this Ph.D. thesis was carried out within the
(P23248) at the University of Music and Performing Arts Vienna, under the
with the international research community during the project time, prelimi-
nary studies were presented and discussed at eligible conferences ranging from
Congress on Acoustics. Reports, which include parts of this thesis, were pub-
lished in the associated proceedings and one study (Chapter 3) was published
under the supervision of others. The footnotes provide explicit details about
Measuring musician-instrument
interaction on single-reed
woodwind instruments
In this chapter commonalities and dierences between the saxophone and the
both instruments are evaluated to prepare the two empirical sound production
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
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
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
wrong adjustments of lower lip pressure from the player inuence the lip-
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
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
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-
Chapter 2. Measuring musician-instrument interaction on single-reed
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
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-
capture tongue-reed interactions, using strain gauge sensors on the reed will
that acts as a pressure-controlled valve to modulate the ow into the instru-
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
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-
8 woodwind instruments
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
too much nger force. Music education encourages loose ngering technique
in this thesis, will gain new insights into the used nger forces of professional
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
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)
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
the missing second harmonic, overblowing produces the sound of the third
Chapter 2. Measuring musician-instrument interaction on single-reed
woodwind instruments 9
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
tem. The main dierence is a thicker instrument body with slightly larger tone
staccato playing) inuence the timing and the sound of the tones.
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).
wind performance, a measurement method that can capture tongue and nger
should not handicap the player, but capture reliable data and be suitable for
10 woodwind instruments
This section describes a method to equip synthetic single reeds with sensors
ing section (2.2) is based on this article. The section also further includes
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
playing condition, traditional reeds have to be wet (Pinard et al., 2003). Dry-
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)
Chapter 2. Measuring musician-instrument interaction on single-reed
woodwind instruments 11
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
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
Synthetic single-reeds gather several properties which favour them over nat-
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-
2.2.2 Method1
Strain gauge sensors
Bonded resistance strain gauges are based on the principle that the resistance of
are produced by a low cost circuit printing technique and function as a sta-
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
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,
12 woodwind instruments
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
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.
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
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
Chapter 2. Measuring musician-instrument interaction on single-reed
woodwind instruments 13
particular dimensions of Bb- clarinet and alto-saxophone reeds. See Figure 2.1
saxophone reed (Légère, strength: 2.25) and a sensor equipped Bb-clarinet reed
For the alto-saxophone, the sensor reed was mounted to a Vandoren (AL3,
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
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).
Chapter 2. Measuring musician-instrument interaction on single-reed
14 woodwind instruments
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
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
woodwind instruments 15
Load cell
Displacement
Strain
gauge
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
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).
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
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).
Chapter 2. Measuring musician-instrument interaction on single-reed
16 woodwind instruments
At this point a deformation of the reed was visually observed during the ex-
the closing point. This linear trend was opposite from our expectations. We
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).
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
woodwind instruments 17
Displacement
Strain
gauge
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
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.
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
18 woodwind instruments
Load cell
Displacement
Strain
gauge
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 5 10 15 20 0 5 10 15 20
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
woodwind instruments 19
2.2.4 Discussion1
This section explained a method to equip synthetic single-reeds with strain
machine, reed displacement and strain gauge signal (reed bending) were mea-
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
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
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
calibration.
This section will focus on the measured eects of articulatory tongue actions
this section have been presented and discussed at two international conferences
2
and were published in the associated proceedings (Hofmann et al., 2012a and
20 woodwind instruments
2.3.1 Introduction 1
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
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
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
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,
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.
1 as published in Hofmann et al. (2012a)
Chapter 2. Measuring musician-instrument interaction on single-reed
woodwind instruments 21
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
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
research groups (van Walstijn and de Sanctis, 2014; Chatziioannou, 2010) used
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
to a probe that ranged into the player's mouth. Second, the reed bending was
Section 2.2. Third, the inner mouthpiece pressure was measured by a small
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-
11.025 kHz (16 bits). For this experiment the mouthpiece was only connected
precise pressure values (in Pascal) inside the mouthpiece and inside the players
22 woodwind instruments
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
Chapter 5).
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.
Chapter 2. Measuring musician-instrument interaction on single-reed
woodwind instruments 23
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:
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.
Chapter 2. Measuring musician-instrument interaction on single-reed
24 woodwind instruments
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
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
woodwind instruments 25
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.
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
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
26 woodwind instruments
1.0
Reed bending
0.5
Mouthpiece pressure
0.0
−0.5
−1.0
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
The blowing pressure signal shows that for tongue-separated tones, the blow-
ing from the player was approximately constant through the note transition
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
veries that the player followed the instructions for the two dierent playing
techniques.
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
Chapter 2. Measuring musician-instrument interaction on single-reed
woodwind instruments 27
a) b)
0.8
0.8
Reed bending
Reed bending
0.4
0.4
0.0
0.0
Mouthpiece pressure (kPa)
4
2
2
0
0
−4
−4
Blowing pressure (kPa)
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).
◦
Looking into the reed bending signal in detail, a 180 phase inversion at
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
Chapter 2. Measuring musician-instrument interaction on single-reed
28 woodwind instruments
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).
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
woodwind instruments 29
1.0
Reed bending
Mouthpiece pressure
0.5
0.0
−1.0 −0.5
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.
Chapter 2. Measuring musician-instrument interaction on single-reed
30 woodwind instruments
the reed during performance and give precise information about tongue-reed
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
(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.
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
separated tones. Dierences also appear in the envelope of the pressure sig-
during the note transition. Observations for the eects of tonguing on the reed
behaviour were similar for saxophone and clarinet and the described charac-
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
the clarinet setup (simple geometry of cylindrical tube) and the enhanced mea-
woodwind instruments 31
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
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
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)
Chapter 2. Measuring musician-instrument interaction on single-reed
32 woodwind instruments
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
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
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.
The acceleration sensors which were attached to the moving parts of the
Chapter 2. Measuring musician-instrument interaction on single-reed
woodwind instruments 33
the force sensor (Key Force 2) did not capture any data (see Figure 2.21).
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
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).
Chapter 2. Measuring musician-instrument interaction on single-reed
34 woodwind instruments
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
woodwind instruments
−1.0 −0.5
TRR
TRC TRC
−8000
TRR
−11000
Reed bending
−14000
5
4
3
2
KB
0
●
0.00 0.05 0.10 0.15 0.20 0.25 0.30
Chapter 2. Measuring musician-instrument interaction on single-reed
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).
woodwind instruments
Chapter 2. Measuring musician-instrument interaction on single-reed
Chapter 2. Measuring musician-instrument interaction on single-reed
woodwind instruments 37
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).
woodwind instruments
Chapter 2. Measuring musician-instrument interaction on single-reed
1.0
0.5
0.0
Sound
woodwind instruments
−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
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.
Chapter 2. Measuring musician-instrument interaction on single-reed
40 woodwind instruments
2.4.3 Discussion
First, a measurement setup to investigate ngerings on the saxophone was
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
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
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
quence, the nger has to be placed on the ring-shaped force sensor developed
will give new insights into the applied nger forces during expressive clarinet
performance.
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
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
Two dierent approaches, to capture nger actions of the player were dis-
Chapter 2. Measuring musician-instrument interaction on single-reed
woodwind instruments 41
cussed. First, the measurement of nger forces to the saxophone keys with
standard force sensors were examined. The recorded data was biased by in-
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-
for future research applications (Hofmann et al., 2013c). Second, nger force
measurements with the sensor clarinet (Weilguni, 2013) were investigated and
A combination of the developed sensor reeds, together with the nger force
42 woodwind instruments
Chapter 3
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
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
3.1 Introduction
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
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.,
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
ticulation does not involve tonguing and its sounding result is the smoothest
note transition. Herby, only changes of the ngerings determine the timing
(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
the player blows constantly. The sound of consecutive portato tones has been
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
Chapter 3. Production and perception of legato, portato and staccato
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
with piano players (Goebl and Palmer, 2008). In woodwind performance, the
with both hands improved temporal stability, compared to tapping with only
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
Perception and action in human motor control are strongly connected. The
based auditory stimuli is based on the ability to recognize related vocal tract
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
Chapter 3. Production and perception of legato, portato and staccato
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
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
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
ger movement directions (pressing for tone onsets versus releasing for tone
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
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
jazz ensembles.
Chapter 3. Production and perception of legato, portato and staccato
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
were given as a score for alto-saxophone (sounding a major sixth lower than
tions. In legato articulation tone repetitions are not possible to play, thus note
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-
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
70 mm
flat side
16.3 mm
4 mm
26 mm
curved side
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.
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
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
Chapter 3. Production and perception of legato, portato and staccato
approximately 1 hour per participant. In total 4644 tones were recorded per
for portato and staccato and 97 tones for legato). After their performances,
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-
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
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 data captured during the experiment contained more than 88,000 played
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
(LDF) based on a wavelet decomposition of the sensor reed signal, where the
The reed signal was decomposed using the Maximal Overlap Discrete Wavelet
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.
Chapter 3. Production and perception of legato, portato and staccato
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
cial treatment of the legato recordings was required to locate tone transitions
MODWT analysis was applied to the legato recordings, but with an adapted
had to be omitted completely from the analysis, as the sensor data indicated
To evaluate the quality of the LDF, it was tested on a small data set which
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-
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.
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).
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
Chapter 3. Production and perception of legato, portato and staccato
annotated landmarks.
Score-performance matching1
However, from the detected landmarks, the correct position of each perfor-
between score and performance data (Gingras and McAdams, 2011; Grachten
et al., 2013). Possible reasons for these dierences can be drastic changes of
ted tones, wrong tones). Additionally, in our case there may be errors from
the landmark detection function (false positives, true negatives) in the per-
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
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
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
information from the sound recordings of the performances were extracted with
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).
Chapter 3. Production and perception of legato, portato and staccato
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
vals (IOI, in ms), as the time interval between two subsequent TRR (onset)
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
value to a sequence played too slow. The temporal precision of the played
sequence, while larger values indicate higher variability in the onset distribu-
tion.
The average signed timing error of all performances during the synchro-
(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
tion (metronome IOI = 250 ms, 178.6 ms, 144.2 ms) and synchronisation
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-
click, participants increased the playing speed in the slow tempo condition
and reduced fast tempi to a more comfortable playing speed (Figure 3.5a,
Chapter 3. Production and perception of legato, portato and staccato
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).
Chapter 3. Production and perception of legato, portato and staccato
−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
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.
other instruments (e.g., piano performance, Goebl and Palmer 2013). The
[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
had to be played with the ngers only (legato), the tongue only (portato note
Chapter 3. Production and perception of legato, portato and staccato
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 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.
restricted our data set to legato and portato recordings, because the onset
playing and compared onset timing between these parts. A two-way repeated
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
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-
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
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.
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
Contrary, tones played only by tonguing (solid line) showed almost a constant
irregularity over all three tempo conditions. This was conrmed by three sepa-
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
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
Chapter 3. Production and perception of legato, portato and staccato
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.
CV values were about the same level as tones played with only ngerings. It
movements under this extreme tempo condition, but did not benet from the
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.
while ascending sequences require ngers to open tone holes. To see if the
direction of nger movements (pressing down versus lifting up) inuences the
depression (Melody 1) to those focussed on lifting the keys (Melody 2). A two-
Tempo (IOI) Slow (245 ms) Medium (178.6 ms) Fast (144.2 ms)
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
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
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
(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
Chapter 3. Production and perception of legato, portato and staccato
0.10
Tongue−reed contact duration (s) ● Portato articulation
● Staccato articulation
0.08
●
0.06
●
0.04
● ● ●
0.02
0.00
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.
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
pendent of the playing speed, while in staccato articulation, the relative gap
duration is constant.
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
indicate whether they felt comfortable when playing the sensor instrument or
not. We tested timing accuracy and timing precision for this group eect by
the two groups, thus, the sensor instrument did not aect the recorded perfor-
Chapter 3. Production and perception of legato, portato and staccato
mances. We also tested for eects of self-reported handedness and skill level
articulation.
3.3.1 Method
Participants
Nineteen female and twelve male (N = 31, mean age = 24 years, range = 19
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
3.2). The group of musicians, who did not play a wind instrument, had a
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
Experimental Design
In a 3 × 3 × 2 × 2 × 2 (3 articulations × 3 intervals × 2 registers × 2
three dierent pitch intervals (major second, minor sixth, minor sixth includ-
ing register change), with legato, portato and staccato articulation, within
Chapter 3. Production and perception of legato, portato and staccato
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
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
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-
back one stimulus. Buttons A and B contained two note transitions played
that matched the articulation used in either A or B. The question our par-
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.
Chapter 3. Production and perception of legato, portato and staccato
The experiment was grouped into two blocks, each containing all 36 stimuli
in four dierent listening orders. Participants were allowed to play back the
block was done the participants lled in a questionnaire about their musical
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.
puted the percentage of wrong answers per participant collapsing across listen-
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-
signicant eect of articulation (F (2, 56) = 67.897, p < 0.001, η 2 = 0.841), but
the legato-portato listening task diered signicantly from the results of the
other two tasks (p < .001). Errors occurred most often when participants
1 A, B and X stimuli contained three dierent articulations: no correct answer was pos-
sible.
Chapter 3. Production and perception of legato, portato and staccato
a) b)
35
20
● Musicians ●
● Musicians
● ● Non−Musicians ● Non−Musicians
30
Saxophonists Saxophonists
●
15
Wrong answers (%)
25
10
●
15
●
●
●
10
5
5
● ●
0
0
legato−portato legato−staccato portato−staccato legato−portato legato−staccato portato−staccato
Articulation 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
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.
the saxophone with two experiments. For the production experiment we built
Chapter 3. Production and perception of legato, portato and staccato
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
with the right arm have to be coordinated with left hand ngerings. Baader
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
from perfect simultaneity (50 ms), but did not lead to audible interruptions.
open multiple tone holes at the same time. This requires simultaneous nger
movements, also called safe nger transitions. Almeida et al. (2009) showed
already lead to audible changes in the radiated sound. Taking these studies
into account, it seems that wind instrumentalists need even more precise nger
tongue.
under dierent tempi, which were produced by dierent eectors (ngers only,
precise than tone onsets produced by ngerings only. Highest precision was
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
Chapter 3. Production and perception of legato, portato and staccato
actions. This nding suggests that ngers play a dominant role in the overall
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
domain. With the help of sensor-equipped wind instruments and the devel-
opment of new customised sensors, useful advice for music education may be
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).
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
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
instrument.
Chapter 3. Production and perception of legato, portato and staccato
selected excerpts from the rst Weber Concerto and a technical exercise (rep-
The clarinet was developed at the Vienna Technical University within the
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
tice for decades until they gain a professional level. Besides studying musical
67
68 Chapter 4. Finger forces in clarinet playing
thumb and index nger, Forssberg et al., 1991), a musician has to hold the
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 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
in violin playing are controlled by right hand bowing techniques, there was
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
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-
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
that during dicult and fast sequences, players used larger nger movements
Palmer et al. (2009b) suggest that there is an eect of tactile feedback on tem-
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
partly supported the multiple-timer model by Ivry et al. (2002). In the case
overall timing. For fast tempi, nger actions showed a dominant inuence on
70 Chapter 4. Finger forces in clarinet playing
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
show similar results as the group of jazz and classical saxophonists. With
applied force to the tone holes and report nger forces used with this technical
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
4.2.1 Methods
Participants
Ten female and thirteen male clarinetists (N = 23, mean age = 26.9 years,
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
Experimental design
For the experiment three dierent performance tasks were prepared: a warm-
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
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
Performing Arts Vienna. Each selection fullled one of three testing condi-
tions (see Appendix B.1), the fourth testing condition will be introduced by
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
Equipment
The following setup was chosen for this experiment: a sensor-equipped Vien-
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-
72 Chapter 4. Finger forces in clarinet playing
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
guni, 2013). The sensor reed, the force sensor rings, the microphone (d:vote
signals simultaneously (A/D conversion with sampling rate 11.025 kHz, 16 bit
Procedure
All participants gave their informed consent to the procedures of the study and
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
the provided metronome signal). For the remaining two trials, the metronome
signal was muted and the instructions were to play expressively (choose own
For a technical exercise (fourth) task, each participant played the 23-tone
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
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
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
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
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
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.
Chapter 4. Finger forces in clarinet playing 75
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
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
5
Ring Sensors
● 1
4
● 2
3
3
●
●
4
● ● ●
●
5
●
6
2
● ● ●
● ● ●
● ●
● ● ● ● ●
● ● ● ● ●
● ● ● ●
●
1
● ● ● ●
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
15
Ring Sensors
● 1
● 2
10
3
● 4
5
●
6
●
5
● ●
● ● ● ● ●
●
● ● ● ●
● ● ●
● ●
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
Chapter 4. Finger forces in clarinet playing 77
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
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
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.
[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.
sequence was performed under two dierent playing instructions from the ex-
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.
78 Chapter 4. Finger forces in clarinet playing
Expression
1.5
1.4
Mean Finger Force (N)
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.
only the trials recorded in the high expression level performance condition are
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
Chapter 4. Finger forces in clarinet playing 79
a) Register b) Dynamics
1.5
1.5
1.4
1.4
Mean Finger Force (N)
1.3
●
●
1.2
1.2
●
●
1.1
1.1
1.0
1.0
0.9
0.9
Low High Soft Loud
c) Tempo
1.5
1.4
Mean Finger Force (N)
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.
80 Chapter 4. Finger forces in clarinet playing
1.8
● Slow Tempo ● Slow Tempo
● Fast tempo ● Fast tempo
1.6
1.6
Mean Finger Force (N)
1.4
●
●
●
1.2
1.2
●
●
●
● ●
1.0
1.0
Low High Soft Loud
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
(dashed line), loud dynamics led to increased nger forces, whereas in fast
a seven-step rating scale (from 3 to +3), whether they think their own nger
Chapter 4. Finger forces in clarinet playing 81
2.5
●
2.0
●
1.5
● ●
●
● ● ●
● ●
● ●
●
1.0
●
●
●
●
●
●
●
0.5
0.0
−2 −1 0 1 2
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
during playing (19%) or during and after playing (24%), an one-way repeated
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
sensor clarinet for practising or for teaching, the results showed no clear trend
the group of players with problems had more interest (70%) in using such a
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
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).
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
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
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
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
students had successfully passed the music university entrance exams which
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
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
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
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
same level as with the saxophone study and showed a similar tempo pattern
Figure 3.6b).
A main dierence in the results of both studies was that the clarinetists'
the slow and the medium tempo condition (clarinetists' mean CV = 0.074,
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.
Chapter 4. Finger forces in clarinet playing 85
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 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-
Separate post-hoc pairwise t-test, three for each tempo condition, conrmed
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
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
Looking at the measured left-hand nger force for the technical exercise task,
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
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
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
precision with nger forces (Fmean ). No such correlation was found in our data
1 To
put a focus on the ngering technique the dataset was restricted to the performances
with the nger-only condition.
Chapter 4. Finger forces in clarinet playing 87
1.0
● Students
● Professionals
0.8
Mean Finger Force (N)
●
●
●
●
0.6
●
●
0.4
0.2
0.0
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
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 =
0.098).
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
● ● ● ●
●
●
IDs
Timing accuracy per participant in technical excercise task
0.2
Tempo condition
● Slow
● Medium
0.1
●
Fast
● ●
● ● ● ●
● ● ●
● ● ● ●
●
0.0
● ●
● ● ●
● ● ● ● ● ● ●
● ● ● ● ● ●
● ● ● ● ●
● ●
●
−0.1
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.
Chapter 4. Finger forces in clarinet playing 89
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 .
usually practise for a longer duration daily (M = 2.78 h, SD = 1.15) than the
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
signicant eect.
4.3 Discussion
This study investigated nger force proles and temporal eects of professional
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
The average nger forces participants used to close the tone holes during
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
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.
lateral stress for the body. Using too much tension during playing may result
tool to pinpoint reasons for overstressed body parts (Grosshauser and Troester,
2013). Although index nger, middle nger and ring nger have not primarily
pothesised that using too much tension with these ngers may be an indicator
ing, did not show signicantly dierent nger force proles compared to the
how much nger force they think they use, they showed good results in their
self-estimation.
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).
Chapter 4. Finger forces in clarinet playing 91
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
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
The results of the study showed that nger forces in clarinet playing are
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
dierent types of clarinets, but also allows to use the setup to capture nger
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
eectors (ngers only, tongue only, tongue-nger actions), under three dier-
trends were found in the results of both studies. For the slow tempo con-
sion in all three tempo conditions, we again observed the eect that for the
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
92 Chapter 4. Finger forces in clarinet playing
Application of performance
sound synthesis
players.
5.1 Introduction
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).
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.,
tones have been developed (Chatziioannou, 2010; Karkar et al., 2012; Van Wal-
changing the control parameters of the model. This allows to modulate the
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-
resonance tube (Chatziioannou and van Walstijn, 2012). During note onsets
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
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
et al. (2009) modelled the tongue as a gate to the mouthpiece, which prevents
2.3.3 and 3.2.2 showed that there is a clear distinction between the signals (reed
synthesis 95
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.
clarinet (straight cylindrical tube). The aim of this study is to model artic-
added to the model. In a rst step, new model parameters for steady-state
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) .
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
Tongue articulation
To model a portato tongue stroke to the vibrating reed, a tongue-reed contact
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.
Chapter 5. Application of performance measurements to physics based sound
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
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.
Tongued tone transition ), the following parameters of the physical model are
adjusted:
ed by the measurements from Chapter 2.3.3 (Figure 2.13, left). The variation
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
synthesis 97
4.0
Blowing pressure (kPa)
3.5
3.0
2.5
2.0
Figure 5.2: Air-separated tones tones are modelled by a modulation of the 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
The presented model parameters make a clear distinction between the two
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
Also for the mouthpiece pressure, a characteristic pressure envelope for each
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
in comparison to the case of air-separated tones. Such an eect has also been
5.3 Discussion
nding the right balance between adding sensor technology to the acoustic
instrument and not distracting the player. Therefore observations are limited
From the obtained blowing pressure, reed bending and mouthpiece pres-
In future work this approach will be enhanced towards the simulation of nu-
the reed bending signal needs to be directly correlated with reed displacement.
may be used as a measurement tool. These tests can range from sound com-
sound samples, this would verify the variation of the model parameters.
Chapter 5. Application of performance measurements to physics based sound
synthesis 99
4e−04
4e−04
Reed displacement (m)
2e−04
0e+00
0e+00
Mouthpiece pressure (kPa)
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).
Reed displacement (m)
2e−04
2e−04
0e+00
0e+00
Mouthpiece pressure (kPa)
1.5
0.5
0.5
−0.5
−0.5
−1.5
−1.5
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).
Chapter 5. Application of performance measurements to physics based sound
100 synthesis
Chapter 6
The research presented in this thesis investigated the interaction between mu-
specic focus was tongue articulation measurements and nger force measure-
tests are presented in Chapter 2. Although the sensor reeds were primarily de-
4). The corpus of data within the production experiments contains more than
101
102 Chapter 6. Conclusion and future work
was developed. After calibrating the reed sensor, it was possible to ex-
stroke shows a damping eect to the vibrating reed, and its force changes
pressure, and reed bending, physical control parameters for tongued and
techniques.
For slow tempi, the performance timing beneted from combined tongue-
nger actions. This supports the multiple-timer model from Ivry et al.
play a dominant role on the timing. Even tough in the case of the clar-
ertheless saxophonists showed the best results for this task. In reference
results of the experiment indicate that the link between production and
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
precise timing in the technical exercise (Fmean = 0.54 N). These observa-
tions indicate that professional players use very light ngering technique
if possible.
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.
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
104 Chapter 6. Conclusion and future work
showed two weaknesses. First, the polymer rings glued on three glass balls fell
the sensing elements and the Low Temperature Co-red Ceramic (LTCC) ring
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
would help to monitor the tip of the sensor reed. The position and the amount
provides details about the link between reed bending measurements and the
tip opening.
cate sound spectra, physical modelling aims to simulate the sound generation
of waveforms and also allows to include player actions that are controlling the
involved physical parameters. With the link between reed bending and tip
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
recorded articulation techniques, this would additionally verify the model pa-
son of style between jazz and classical soloists. Applying stylistically grouped
video technology, and acceleration sensors, also on the musician's body, gives
from large pick-up gestures to the ne motor control of each individual player.
106 Chapter 6. Conclusion and future work
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
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,
107
108 BIBLIOGRAPHY
Bergeot, B., Almeida, A., Vergez, C., and Gazengel, B. ( 2012). Attack tran-
sients in a clarinet model with time-varying blowing pressure, in Proceedings
(Budapest, Hungary).
im Rahmen der 30. Tage Alter Musik in Herne 2005, chapter 'Zur Geschichte
114.
1189.
29.
629639.
Chen, J., Smith, J., and Wolfe, J. ( 2011). Saxophonists tune vocal tract reso-
nances in advanced performance techniques, The Journal of the Acoustical
Dahl, S. (2004). Playing the accent - comparing striking velocity and timing
in an ostinato rhythm performed by four drummers, Acta Acustica united
Diethelm, U. 2011).
( Schmerzen beim Musizieren, Schweizerische
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).
Fletcher, N. H. ( 1979). Air ow and sound generation in musical wind in-
Gazengel, B., Guimezanes, T., Dalmont, J.-P., Doc, J., Fagart, S., and Léveillé,
ing using both structural and temporal information from midi recordings,
holm Music Acoustics Conference 2013, SMAC 13, edited by R. Bresin and
Sweden).
Guillemain, P., Vergez, C., Ferrand, D., and Farcy, A. ( 2010). An instru-
Publishing).
Hofmann, A., Chatziioannou, V., Kausel, W., Goebl, W., Weilguni, M., and
Hofmann, A., Chatziioannou, V., Weilguni, M., Goebl, W., and Kausel, W.
Stockholm, Sweden).
Hofmann, A., Goebl, W., Weilguni, M., Mayer, A., and Smetana, W. ( 2012b).
Measuring tongue and nger coordination in saxophone performance, in
and 8th Triennial Conference of the European Society for the Cognitive Sci-
Hofmann, A., Goebl, W., Weilguni, M., and Smetana, W. ( 2013c). Zooming
into saxophone performance: Tongue and nger coordination, in Proceed-
Hori, M., Hosono, M., Takahashi, H., Matsumoto, K., and Shimoyama, I.
Ivry, R., Richardson, T., and Helmuth, L. ( 2002). Improved temporal sta-
Jensenius, A., Wanderley, M., Godøy, R., and Leman, M. ( 2010). Musical Ges-
tures: Sound, Movement, and Meaning, chapter Musical gestures: concepts
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
8th Triennial Conference of the European Society for the Cognitive Sciences
Patent 6,087,571.
Legere, G., Kortschot, M., et al. ( 2011). Oriented polymer reeds for woodwind
instruments, US Patent 7,902,443.
meaning, chapter 'Why Study Musical Gestures?', 311 (New York: Rout-
ledge).
Publications).
S. N. F., Gerwig, M., Habas, C., Hagura, N., Ivry, R. B., Mariën, P., et al.
11, 457487.
Music).
Palmer, C., Koopmans, E., Carter, C., Loehr, J., and Wanderley, M. ( 2009a).
Synchronization of motion and timing in clarinet performance, in Interna-
Palmer, C., Koopmans, E., Loehr, J., and Carter, C. ( 2009b). Movement-
Press).
Percival, D. and Walden, A. ( 2006). Wavelet Methods for Time Series Anal-
ysis, Cambridge Series In Statistical And Probabilistic Mathematics (Cam-
Pinard, F., Laine, B., and Vach, H. ( 2003). Musical quality assessment of
clarinet reeds using optical holography, The Journal of the Acoustical So-
41, 312.
Repp, B. H. ( 2005). Sensorimotor synchronization: a review of the tapping
literature, Psychonomic Bulletin & Review 12, 969992.
Press).
(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
Tzanetakis, G., Essl, G., and Cook, P. ( 2001). Audio analysis using the
Wanderley, M., Vines, B. W., Middleton, N., Mckay, C., and Hatch, W.
Weilguni, M., Hofmann, A., Smetana, W., and Goebl, W. ( 2013). Fabri-
force sensors used for performance studies in clarinet playing, in The 17th
Weilguni, M., Smetana, W., Hofmann, A., Radosavljevic, G., Goebl, W., and
Taiwan).
experiments
119
120 Chapter A. Additional material saxophone experiments
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
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
0
1 1
15
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
0.970.980.970.91 0 0.910.990.980.970.99 1 1 1 1 1 1
5
0 0.710.890.950.970.98 1 1 1 1 1 1 1 1 1 1 ● ● ●
5 10 15 −3 −2 −1 0
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
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
0
1 1
15
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
0.970.980.91 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
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
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).
Low Register C A B H
High Register E F G D
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.
126 Chapter B. Additional material clarinet experiments
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.
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.
Chapter B. Additional material clarinet experiments 127
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.
128 Chapter B. Additional material clarinet experiments
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 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).
Chapter B. Additional material clarinet experiments 129
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).
130 Chapter B. Additional material clarinet experiments
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.
Chapter B. Additional material clarinet experiments 131
Figure B.9: Left-hand index nger force (in Newton) on clarinet, captured for par-
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.
132 Chapter B. Additional material clarinet experiments
Figure B.10: Left-hand index nger force (in Newton) on clarinet, captured for par-
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.
Chapter B. Additional material clarinet experiments 133
134 Chapter B. Additional material clarinet experiments