Influencia de Personas Con Dif Edades y Cultura Usanod Auris

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Audio Engineering Society

Convention Paper 9177


Presented at the 137th Convention
2014 October 9–12 Los Angeles, USA
This Convention paper was selected based on a submitted abstract and 750-word precis that have been peer reviewed by at least
two qualified anonymous reviewers. The complete manuscript was not peer reviewed. This convention paper has been repro-
duced from the author's advance manuscript without editing, corrections, or consideration by the Review Board. The AES takes
no responsibility for the contents. Additional papers may be obtained by sending request and remittance to Audio Engineering
Society, 60 East 42nd Street, New York, New York 10165-2520, USA; also see www.aes.org. All rights reserved. Reproduction of
this paper, or any portion thereof, is not permitted without direct permission from the Journal of the Audio Engineering Society

The Influence of Listeners’ Experience, Age,


and Culture on Headphone Sound Quality
Preferences
1 2 3
Sean E. Olive , Todd Welti , and Elisabeth McMullin

Harman International, Northridge, CA, 91329, USA

1
Sean.Olive@harman.com
2
Todd.Welti@harman.com
3
Elisabeth.McMullin@harman.com

ABSTRACT

Double-blind headphone listening tests were conducted in four different countries (Canada, USA, China and
Germany) involving 238 listeners of different ages, gender and listening experiences. Listeners gave comparative
preference ratings for three popular headphones and a new reference headphone that were virtually presented
through a common replicator headphone equalized to match their measured frequency responses. In this way, biases
related to headphone brand, price, visual appearance and comfort were removed from listeners’ judgment of sound
quality. On average, listeners preferred the reference headphone that was based on the in-room frequency response
of an accurate loudspeaker calibrated in a reference listening room. This was generally true regardless of the
listener’s experience, age, gender and culture. This new evidence suggests a headphone standard based on this new
target response would satisfy the tastes of most listeners.

music and infotainment wherever they go. Yet, the


science related to the perception and measurement of
1. INTRODUCTION headphone sound quality is not well understood. There
appears to be little consensus among headphone
Headphone sales reached $8.2 billion worldwide in manufacturers on how to optimize the design of a
2013, and they continue to grow [1]. The popularity of headphone for good sound. This situation hasn’t
headphones is tied to the growth of smartphones and changed much since 2001 when Poldy summed up the
other mobile devices that allow consumers to enjoy headphone industry as follows:
Olive et al. Headphone Sound Quality Preferences

“A palate of different frequency responses is available experiences participated. The listeners gave comparative
to cater for individual preferences, and each preference ratings for four headphones including three
manufacturer has its own headphone philosophy with popular models (Sennheiser HD800, Audeze LCD-2,
frequency responses ranging from flat to free field and and Beats Studio Edition), and a headphone calibrated
beyond” [2]. to a new target response reported in [6]. A virtual
headphone listening test method was used in order to
Researchers have recently shown experimental evidence remove non-auditory biases (brand, price, visual
that headphones calibrated to the diffuse-field (DF) and appearance, comfort, celebrity endorsement, etc.) from
free-field (FF) target responses are less preferred to listeners’ judgment of the headphone sound quality [10].
other target response curves. Lorho recommended a
modified DF target that reduced the 3 kHz peak from 10 The organization of this paper is as follows: Section 2
dB to 3 dB [3]. Fleischmann et al. found listeners describes the method and test procedure of the
preferred a headphone target response, which more headphone listening experiments, Section 3 describes
closely matched the sound of a loudspeaker in an ITU-R the results of the listening tests, with discussion and
listening room [4]. However, they never divulged the conclusions in sections 4 and 5.
details on what the headphone target response curve
should look like. More recently, Olive et al. provided
experimental evidence that showed a headphone target 2. EXPERIMENT ON HEADPHONE SOUND
response based on in-room measurements of an accurate PREFERENCE
loudspeaker calibrated in a reference listening room was
preferred to DF, FF and other target responses of This section describes the method, test setup and
headphones highly regarded for their sound quality [5], procedure employed for the listening tests.
[6]. The notion that headphones should simulate the in-
room response of an accurate loudspeaker makes perfect 2.1. Selection of Headphones
sense given that most stereo recordings are mixed and
optimized to sound good through loudspeakers in Three popular headphones were selected for the
rooms. Most listening rooms provide bass reinforcement listening tests listed in Table 1.
from room modes and boundary effects, yet headphones Table 1 The prices and descriptions of the headphones
calibrated to the DF and FF target response curves lack models used in this study.
this “room gain”, and therefore sound too bright and
thin.
Headphone Brand & Price Description
Together, these four recent studies provide scientific Model
evidence and the motivation to develop a better industry
headphone standard that promotes optimal headphone Sennheiser HD800 $1500. Over-ear (OE), open
sound quality. However, a new headphone standard will back (OB) with ring
not likely gain wide acceptance unless there is evidence radiator driver
it satisfies the tastes of a broad demographic based on
listeners’ age, listening experience and culture. Previous Audeze LCD2 $995. OE and OB with a
studies on loudspeaker sound quality preferences [7], version 2 magnetic planar driver
[8] have shown that listeners tend to prefer the most
accurate models regardless of listening experience and Beats by Dre Studio $299 OE, closed back with
culture. While a recent headphone study [9] suggests Edition dynamic driver and
that college students and trained listeners have similar active noise cancellation
headphone preferences, the effects of age, training and
culture on headphone sound preferences have not been Harman Headphone $119 OE, OB, dynamic driver
thoroughly investigated. Target Response equalized to Harman
implemented on Target Response [6]
To that end, this paper reports the results of a series of replicator phone:
double blind listening tests conducted in four countries Sennheiser HD518
(Canada, United States, China and Germany) where 238
listeners from a broad range of age and listening

AES 137th Convention, Los Angeles, USA, 2014 October 9–12


Page 2 of 11
Olive et al. Headphone Sound Quality Preferences

The headphones ranged in price from $119 to $1500. Fig. 1 shows the measured frequency responses of the
The Sennheiser HD800 was an open-back model physical headphones compared to the virtual versions.
employing a ring radiator driver. The Audeze LCD2
was an open back headphone using a magnetic planar
driver. Both models are considered to be among the best
sounding headphones according to various headphone
review and enthusiast websites. The third headphone
was a Beats Studio Edition headphone, a closed-back (a)
headphone with a dynamic driver and active noise
control (ANC). The fourth headphone was the
Sennheiser HD518, which was an open back design
with a dynamic driver. This headphone was also used as
the replicator headphone.

2.2. Virtual Headphone Method

A virtual headphone method was employed in order to


control and remove non-auditory biases (headphone
brand, price, visual appearance, celebrity endorsement, (b)
fit and comfort) from the listeners’ judgment of sound
quality. In a previous paper [10], we tested and
validated the virtual headphone method by comparing
listeners’ preference and spectral balance ratings of six
different headphones using both virtual and standard
test methods where the actual physical headphones were
compared. The correlation in headphone preference
ratings between virtual and standard test methods was
quite high (r = 0.85). We speculated that higher
correlations could be achieved if headphone leakage
(c)
effects were better simulated for individual listeners,
and biases related to headphone weight/comfort were
removed from the standard method.

2.2.1. Measurement and Equalization of


Headphones

The frequency response of the left and right channels of


each headphone was measured on a GRAS 45 AG
coupler equipped with the adult ITU type 3.3 pinna. An
average of nine measurements was taken based on nine (d)
reseats of the headphone on the coupler to minimize
errors related to how the headphone was positioned on
the coupler.

Using the auto-EQ function of Harman’s Audio Test


System (HATS), an infinite impulse response filter was
Figure 1 The left (upper black curves) and right (lower
designed to simulate the frequency response of each
set of red curves) channel frequency responses of the
headphone on the simulator headphone. The virtual
actual (dotted curves) and virtualized (solid curves)
filters were then implemented on the playback device
headphones for the (a) Harman target curve (b)
(Apple iPad 4th Generation) using Apple’s Core Audio
Sennheiser HD800 (c) Audeze LCD2 and (d) the Beats
frameworks.
by Dre Studio Edition.

AES 137th Convention, Los Angeles, USA, 2014 October 9–12


Page 3 of 11
Olive et al. Headphone Sound Quality Preferences

The graphs indicate that there is very good agreement participated in this study ranging in age from 20 to 70
between the physical versus virtual headphone up to years age (median age = 29 years, standard deviation =
about 10 kHz. We chose not to equalize the headphone 11 years). Seven different listening groups were tested
below 30 Hz (to avoid low frequency distortion) and at five locations in four countries: USA, Canada,
above 10 kHz due to measurement errors related to Germany and China. Four of the listening groups were
headphone position on the coupler. compromised of Harman employees from different
locations in United States, Germany and China. The
2.3. Selection of Listeners other three listening groups were made up of
undergraduate college students enrolled in audio
Table 2 summarizes details about the different engineering/production programs: Citrus College and
participants in this study related to their nationality, age, Loyola Marymount University (LMU) both located in
gender, and listening experience. A total of 238 listeners the Los Angeles area, and Harris Institute located in
Toronto, Canada.

Table 2 Details on the different listening groups used in this study.

Group Country Count Median Age Gender Experience Description


(Standard Count N = None
Deviation) L = A Little
M = Medium
AL = A Lot

Harman NR USA 9 39 (10.9) Male = 7 N (0) L (1) M (2) AL(6) Harman employees
Female = 2 in Northridge,
California

Harman FH USA 23 38 (13.4) Male = 20 N (6) L (3) M (4) AL(10) Harman employees
Female = 3 in Farmington Hills,
Michigan, USA

Harman KB Germany 72 38 (9.2) Male = 67 N (24) L (19) M (18) AL (11) Harman employees
Female = 5 in Karlsbad,
Germany

Harman SZ China 26 31 (6.5) Male = 19 N (3) L (10) M (1) AL (12) Harman employees
Female = 7 in Shenzhen, China

Citrus USA 24 23 (5.5) Male = 18 N (5) L (6) M (9) AL (4) Audio students at
College Female = 6 Citrus College in
Los Angeles, USA

LMU USA 15 21 (1.2 Male = 14 N (0) L (3) M (12) AL (0) Audio students of
Female = 1 Loyola Marymount
University in Los
Angeles, USA

Harris Canada 69 23 (8.9) Male = 60 N (5) L (11) M (20) AL (33) Audio students and
Institute Female = 9 AES members in
Toronto, Canada

AES 137th Convention, Los Angeles, USA, 2014 October 9–12


Page 4 of 11
Olive et al. Headphone Sound Quality Preferences

The gender makeup of the listening panel was 86% SD – Steely Dan / Two Jazz Pop with Male
male and 14% female. Against Nature / Cousin Vocal
Dupree / Giant Records
Prior to the test, each listener was asked to describe /B00004GOXS/ February 29
their critical listening experience using one of the four 2000
possible categories. The four listening experience
categories and the percentage of listeners who chose DD / Dido/ Safe Trip Home Rock Pop with
them were: none (18%), a little (22%), a medium /Don’t Believe in Love/ Arista/ Female Vocal
amount (23%), and a lot of experience (37%). Included B001EO2UKO / November 18
in this last category were Harman trained listeners (44% 2008
of the category) from USA, Germany and China. These
39 listeners all had tested normal audiometric hearing,
had successfully passed the Harman listener training (a
minimum of level 7 of Harman’s How to Listen [11]), 2.5. iPad Listening Test Software
and participated in formal listening tests on sound
quality evaluation. A custom listening test application was written for
Apple iPad to administer the tests in an automated and
The remaining participants were not tested for normal repeatable manner. The software application
audiometric hearing, nor were they formally tested for administered all aspects of the test including
their critical listening ability. A common feature among presentation of treatments, and collection and storage of
the 238 listeners is that they had an interest in audio. the listeners’ responses. It also provided the equalization
The listeners either worked for an audio manufacturing to simulate the frequency responses of the different
company or they were currently students enrolled in an headphones under test (see section 2.2.1).
audio production course, and considering a career in this
area. Fig. 2 shows the graphical user interface of the listening
test application.
2.4. Program Selections

Table 3 summarizes the three music selections used in


these listening tests. The tracks were digitally copied
from commercial CD, and edited into short 15-20 s
loops. All three tracks were spectrally analyzed and
deemed to have sufficiently dense and broadband
spectra to facilitate listeners’ judgment of potential
differences among the headphones in terms of their
spectral balance, bandwidth extension and coloration.

Table 3 Details on the three program selections used in


this study.

Program ID /Artist / CD / Description


Track Name/ Record Figure 2 A screen shot of the custom iPad application
Label/ASIN/ Release Date used for these listening tests.

ES / Estelle w. Kayne West / Hip Hop with male The graphical buttons (A through D) allowed the
Shine/ American Boy / Atlantic and female vocal listener to switch among the different headphone
Records, 2008, B00142Q7H8 / treatments or pause the music. The four sliders below
April 29 2008 each button allowed listeners to enter their numerical
preference ratings for each headphone. The 11-interval

AES 137th Convention, Los Angeles, USA, 2014 October 9–12


Page 5 of 11
Olive et al. Headphone Sound Quality Preferences

preference scale was defined from 0 to 10 with semantic 3.1. Statistical Analysis
descriptors at every second interval: 1 (Strongly
dislike), 3 (Dislike), 5 (OK - Neither Like Nor Dislike), The results were analyzed using 5-way repeated
7 (Like), and 9 (Strongly Like). To encourage a measures analysis of variance (ANOVA). The within-
consistent use of the scale, listeners were given the subject fixed factors were Headphone (4 levels),
following guidelines in choosing how they separated Program (3 levels) and Observation (2 levels). The
their ratings between two headphones: a ≤ 0.5 point between-subject factors were Listening Group (7 levels)
separation implied a slight preference, 1 point and Listening Experience (4 levels). The dependent
separation a moderate preference, and ≥ 2 points variable was preference rating. All statistical tests were
separation implied a strong preference. Listeners could performed at a significance level of 5%.
optionally provide comments to describe the underlying
reasons for their preferences. The software application Prior to ANOVA, inspection of the residuals confirmed
automatically randomized the order in which programs the underlying assumptions of ANOVA were not
and headphones were presented in each trial. violated. As a precaution, nonparametric analysis of the
data was performed and produced the same effects and
2.6. Relative and Absolute Playback Levels conclusions as the ANOVA. Therefore, we only report
the results of the ANOVA in the sections below.
The relative playback levels of the different headphones
were matched to the same loudness level according to The main effects on preference were Headphone: F (3,
the ITU-R 1770.2 loudness meter. The average absolute 702) = 498.17, p = <0.001; Program: F (2, 468) = 4.9, p
playback level was set to a comfortable level: 78 dB (B- = 0.008; and Listening Experience: F (3, 234) = 3.4, p =
weighted slow). All the tests were conducted in quiet 0.02. There also were significant interactions between
listening rooms to minimize the influence of Headphone * Program: F (6, 1404) = 17, p < 0.001; and
background noise on listeners’ headphone preferences. Headphone * Observation: F (3, 702) = 3.2, p = 0.02.

2.7. Listening Test Procedure 3.2. Effect of Headphone on Preference

Prior to the test, listeners were given listening test Fig. 3 plots the mean preference ratings and 95%
instructions including a demonstration of the test confidence intervals based on all 238 listeners. A post-
software so that they were familiar with how to switch hoc Scheffé test confirmed that the differences in
between headphones and enter their ratings and preference among all headphones were statistically
comments. Listeners were told they would be evaluating significant.
different headphone equalizations and rating them based
on their personal preferences in sound quality.

Each listener completed a total of six trials using three


different programs with two observations. There were
no time limits imposed on the listener to complete the
test. On average, the test was completed in about 15
minutes. Listeners could pause the test at any time to
take a break, or ask the test administrator for help
regarding the test instructions and software user
interface.

3. RESULTS

Section 3 presents the statistical analysis of the Figure 3 The mean preference ratings and 95%
headphone preference listening tests and the main confidence intervals for the headphone treatments based
effects. on all 238 listeners.

AES 137th Convention, Los Angeles, USA, 2014 October 9–12


Page 6 of 11
Olive et al. Headphone Sound Quality Preferences

Note that from this point on, headphones are labeled as


HP1 through HP4 based on the order in which they were
preferred. The most preferred headphone was based on
the Harman target curve, HP1 (preference rating = 6.2).
It was slight-moderately preferred to HP2 (5.8),
moderate-strongly preferred to HP3 (5.0), and strongly
preferred to HP4 (2.2). Both HP2 and HP3 were also
strongly preferred to HP4. HP2 was slightly-moderately
preferred to HP3.

3.3. Effect of Program on Preference

Program had a small effect on preference ratings (see


Fig. 4). The effect was less than 0.2 of a preference
rating. On average, the headphones received higher
preference ratings for program P1 (“American Boy”)
relative to programs P2 (“Cousin Dupree”) and P3
(“Don’t Believe in Love).
Figure 5 The mean preference ratings and 95%
confidence intervals for the interaction between the
factors Program and Headphone.

3.5. Effect of Listening Experience on


Headphone Preference

Listening experience had a significant effect on how the


listeners used the preference scale (see Fig. 6).

Figure 4 The mean preference ratings and 95%


confidence intervals showing the effect of Program
averaged across all headphones and listeners.

3.4. Interaction Effect Between Program and


Headphone

The significant interaction effect between Program and


Headphone is graphically illustrated in Fig. 5. The effect
was largely isolated to Program P3 (Dido’s “Don’t
Believe in Love”), which produced slightly higher
preference ratings for the HP1, and slightly lower Figure 6 The mean preference ratings and 95%
ratings for HP3. Apart from this interaction, the confidence intervals based on the listeners’ self-assessed
headphone ratings were relatively independent of the listening experience.
Program.
The less experienced listeners tended to give higher
preference ratings than the more experienced listeners.
This behavior is consistent with previous studies that
examined the influence of listener training on
loudspeaker [7],[8],[12] and headphone preference

AES 137th Convention, Los Angeles, USA, 2014 October 9–12


Page 7 of 11
Olive et al. Headphone Sound Quality Preferences

ratings [9]. In those studies, trained listeners tended to between-subjects factor Country (4 levels: Canada,
give lower, more reliable and discriminating preference USA, Germany, China), and the fixed within-subject
ratings than the untrained listeners. In those studies, factors Headphone (4 levels), Program (3 levels), and
both trained and untrained listeners generally preferred Observation (2 levels). The dependent variable was
the same loudspeakers and headphones. The next preference rating.
section addresses whether the listeners in the current
study also preferred the same headphones regardless of Country had a significant effect on preference: F (3,
their prior listening experience or training. 232) = 5.64, p = 0.001. There was also a significant
interaction between Country and Headphone: F (9,696)
= 2.34; p < 0.012. These effects are discussed in the
3.6. Interaction Effect Between Listening following sections.
Experience and Headphone
3.8. Effect of Listeners’ Country of Residence
Fig. 7 plots the mean headphone preference ratings for on Preference
the four different listener groups categorized according
to their self-assessed listening experience. The first Fig. 8 summarizes the mean preference ratings and 95%
observation is listeners generally preferred the same confidence intervals based on the listeners’ country of
headphones based on ranking, regardless of their residence. The average preference ratings from listeners
training or listening experience. in Canada and China were identical, and slightly higher
than those given by listeners in Germany and USA. This
could likely be explained by differences in the makeup
of trained and untrained listeners among the four
countries we tested. For example, there were no Harman
trained listeners in the Canada sample, which could
explain the higher ratings. Additional analysis could
confirm or deny if this explanation was valid.

5.5
Preference Rating

5.0
Figure 7 The mean headphone preference ratings and
95% confidence intervals for different listener
categories based on their listening experience. 5.08 5.08
4.5
4.86
However, there was an interaction effect between 4.48
Listening Experience and Headphone related to how the
listeners scaled their preference ratings. The listeners 4.0
Canada China Germany USA
with more experience, denoted in Fig. 7 as “Medium
Listeners’ Country
Amount” and “A Lot”, were able to discriminate more
among the four headphones, as indicated by the wider Figure 8 The mean preference ratings and 95%
range of scores given to the them. The listeners with confidence intervals for each headphone based on the
less experienced, denoted as “None” and “A Little” country of residence of the listener.
tended to assign higher ratings to the headphones, and
were less able to discriminate between HP1 and HP2.
3.9. Interaction Effect Between Country and
3.7. Does Listener Nationality Influence Headphone
Headphone Preference?
A more important question in this study was whether
To answer this question the data was analyzed using a the listeners’ country of residence or culture influenced
repeated measures ANOVA model that included the their headphone preferences. Fig. 9 shows the mean

AES 137th Convention, Los Angeles, USA, 2014 October 9–12


Page 8 of 11
Olive et al. Headphone Sound Quality Preferences

preference ratings and 95% confidence intervals for the ANOVA model included Age (5 levels) as a between-
different headphones according to the listeners’ country subjects factor, and found no significant effect on
of residence. preference due to age; F (4,232) = 0.24, p = 0.45.

The first observation is that the listeners’ country had There were also no interaction effects between Age and
little influence on their headphone preferences. The Headphone on preference as illustrated in Fig. 10.
order in which the headphones were rated from most Headphone preferences were generally the same from
preferred to least preferred was the same for Canadians, the youngest to the oldest age categories with one noted
Chinese, Germans and Americans. The main differences exception: the oldest listeners (> 55 years) were not able
among these listening groups are how they weigh their to discriminate between HP1 and HP2. These listeners
likes-dislikes for certain headphones. The Americans also had the largest error bars, which is related to the
and Canadians tended to rate HP4 lower than the other relatively low number of listeners (i.e. eight) in this
listeners, with the Chinese being the most forgiving of category. Such a small sample of older listeners makes
this headphone. The Germans gave slightly higher it difficult to draw any general conclusions about their
ratings to HP2 than did the other groups. headphone preferences.

Headphone: HP1 HP2 HP3 HP4

6
Preference Rating

1
Canada China Germany USA
Listeners’ Country

Figure 9 The mean preference ratings and 95% Figure 10 The mean preference ratings and 95%
confidence intervals for the different headphones shown confidence intervals for different age categories of
as a function of the listeners’ country of residence. listeners.

Finally, it is important to note that the effects of hearing


3.10. Does the Listeners’ Age Influence Their loss (noise-induced and age-related loss) were not
Headphone Preferences? factored into this study. Only the trained Harman
listeners from US, Germany and China were tested for
For decades, many audio marketing departments have normal audiometric hearing. This could be important
posited that audio playback systems require different since Toole has reported hearing loss adversely affects
sonic signatures to satisfy the alleged different tastes of listener performance and loudspeaker preferences [13].
different demographic groups defined according to age,
culture and socio-economic status. To date, there have 4. DISCUSSION
been no published studies to support this rationale, yet
the belief still persists.
In the previous section, we presented experimental
evidence that showed listeners, on average, preferred
To determine whether the age of the listener influenced the same headphones regardless of their listening
their headphone preferences, we categorized listeners experience, age or country of residence (i.e. Canada,
into the following five age categories in years: < 25, 25- USA, China and Germany). The most preferred
34, 35-44, 45-54, and > 55. A repeated measures headphone (HP1) was based on the Harman target

AES 137th Convention, Los Angeles, USA, 2014 October 9–12


Page 9 of 11
Olive et al. Headphone Sound Quality Preferences

response that simulates the in-room response of an listeners’ judgment of sound quality. The listeners
accurate loudspeaker calibrated in a reference listening included both trained and untrained listeners over a
room. This study confirms the results from previous wide range of age groups from Canada, USA, China and
studies [5], [6] where this target response was preferred Germany. This allowed us to study the influence of age,
to other choices. gender, listening experience and nationality on
headphone sound preferences Based on the statistical
Looking at Fig. 1, there is a clear relationship between evidence presented in section 3, there are three main
the measured frequency responses of the headphones conclusions:
and how listeners rated them based on their sound
quality preferences. The most preferred headphone 1. Listeners generally preferred the same
(HP1) has the measured response shown in Fig. 1 (a). headphones regardless of their listening
The other headphones (HP2, HP3 and HP4) received experience, age or country of residence.
preference ratings that tended to correspond to how far
their response deviated from the response of HP1. The 2. As a group, the experienced listeners gave
target response of HP1 could be the basis for a new lower and more discriminating headphone
headphone standard that would likely have wide preference ratings than the less experienced
acceptance among listeners of different age groups, listeners.
cultures, and listening experiences.
3. The most preferred headphone (HP1) had a
This study also confirms findings from previous studies measured target response that simulated the in-
in which the authors reported that listeners generally room response of an accurate loudspeaker
preferred the most spectrally well-balanced calibrated in a reference listening room. The
loudspeakers [7], [8], [13] and headphones [5], [6], [9], more the frequency response of the other
[10] regardless of age, or listening experience. In [8], headphones deviated from this, the less they
the author reported that Japanese and American college were preferred.
students preferred the same loudspeakers as trained
American listeners. These studies along with the current It is important not to generalize the results of this study
one suggest that “good sound” is more of a universal beyond the conditions tested here. Ideally, a much larger
percept rather than a matter of personal taste influenced sample of listeners from different countries should be
by culture, age or experience. tested before claiming that cross-cultural tastes in
headphone sound either exist or don’t exist. Hopefully,
In this current study, listening experience had an this study will encourage other researchers to continue
expected effect on preference (see Fig. 6) that has been this work. Based on these results, a new headphone
reported in previous studies [7]-[9]. Listeners with more standard based on the target response of HP1 would
experience and training tended to give more likely be widely accepted among listeners. It is
discriminating and lower preference ratings than encouraging to find further experimental evidence that
listeners with less experience. However, as previous there is a wide acceptance among listeners for accurate
related studies have shown, the headphone preferences headphone reproduction, as there is for accurate
(see Fig. 7) were generally the same regardless of loudspeaker reproduction.
listening experience. Apparently, humans are quite
adept at recognizing good sound from bad sound when
given the opportunity to compare them under controlled 6. ACKNOWLEDGEMENTS
listening conditions.
Harman International supported this work. The authors
would like to thank Dr. Rafael Kassier, Steven Li, and
5. CONCLUSIONS Jonathan Pierce from Harman who recruited the
listeners and assisted in administering the tests in
This paper reported the results of a series of controlled Harman Germany, China, and Michigan. We are
double blind listening tests where 238 listeners gave indebted to Blair Francey at Harris Institute in Toronto
comparative preference ratings for four headphones for recruiting the listeners and allowing us to use their
using a virtual headphone method. In this way, non- facilities for testing. We also thank Linda Gedemer
auditory biases (e.g. headphone brand, price, celebrity (Loyola Marymount University) and Stephen O’Hara
endorsement, weight, comfort, etc.) were removed from (Citrus College) for providing college students for these

AES 137th Convention, Los Angeles, USA, 2014 October 9–12


Page 10 of 11
Olive et al. Headphone Sound Quality Preferences

tests. Finally, we thank all of the listeners who Symposium of Electroacoustic Technology
participated in these tests. (ISEAT), Shenzhen, China, (2013).

[10] Olive S.E., Welti T., and McMullin E., “A Virtual


7. REFERENCES
Headphone Listening Test Methododology,”
presented at the 51st Audio Eng. Soc. International
[1] Future Source Consulting, “Headphone Market to Conference, Helsinki, Finland, (August 2013).
Exceed $8 Billion in 2013, www.futuresource-
consulting.com/2013-11-headphones.html (2013 [11] Harman How to Listen: A Listener Training
November). Software Program For Mac and Windows,
www.harmanhowtolisten.blogspot.com (July 2014).
[2] Poldy C. A., “Headphones,” in J. Borwick, Ed.,
Loudspeaker and Headphone Handbook, 3rd
[12] Bech, Soren, “ Selection and Training of Subjects
Edition, pp. 585-692 (Focal Press, Oxford, 2001).
for Listening Tests on Sound-Reproducing
Equipment,” J. Audio Eng. Soc., vol. 40, pp. 590-
[3] Lorho Gaëtan, “Subjective Evaluation of 610 (1992 July/August).
Headphone Target Frequency Responses”, J. Audio
Eng. Soc., preprint 7770, (2009 May).
[13] Toole, Floyd E. Sound Reproduction: The
Acoustics and Psychoacoustics of Loudspeakers
[4] Fleischmann, F., Silzle, Andreas, and Plogsties, and Rooms, Focal Press (2008),
Jan, “Identification and Evaluation of Target
Curves for Headphones”, presented at the 133rd
Audio Eng. Soc. Convention, preprint 8740, (Oct.
2012).

[5] Olive Sean E., Welti Todd, and McMullin


Elisabeth, ”Listener Preference For Different
Headphone Target Response Curves,” presented at
the 13th Convention, Audio Eng. Soc., preprint no.
8867, (2013 May).

[6] Olive Sean E., Welti Todd, and McMullin


Elisabeth, ”Listener Preferences for In-Room
Loudspeaker and Headphone Target Responses,”
presented at the 13th Convention, Audio Eng. Soc.,
preprint no. 8867, (2013 Sept.).

[7] Olive Sean E, “Differences in Performance and


Preferences of Trained versus Untrained Listeners
in Loudspeaker Tests: A Case Study,” J. Audio
Eng. Soc., vol. 51, No. 9, pp. 806-825 (2003
September).

[8] Olive Sean E, “Some New Evidence That


Teenagers and College Students May Prefer
Accurate Sound Reproduction,” presented at the
133nd AES Convention, preprint 8683, (2012
April).

[9] Olive Sean E.,” Do College Students Prefer the


Same Headphone Sound Quality as Trained
Listeners?” presented at the International

AES 137th Convention, Los Angeles, USA, 2014 October 9–12


Page 11 of 11

You might also like