Audio Engineering Society

Convention Express Paper 79

Presented at the 154th Convention
2023 May 13–15, Espoo, Helsinki, Finland

This Express Paper was selected on the basis of a submitted synopsis that has been peer reviewed by at least two qualified
anonymous reviewers. The complete manuscript was not peer reviewed. This express paper has been reproduced from the
author’s advance manuscript without editing, corrections, or consideration by the Review Board. The AES takes no responsibility
for the contents. This paper is available in the AES E-Library (http://www.aes.org/e-lib), 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.

Auditory localisation of low-frequency sound sources

Madalina Nastasa1 , Ville Pulkki1 , and Aki Mäkivirta2
1 Aalto University, Department of Signal Processing and Acoustics
2 Genelec Oy, Olvitie 5, Iisalmi, Finland
Correspondence should be addressed to Madalina Nastasa ([email protected])

ABSTRACT
It is generally thought that humans cannot detect the direction of sound in the very low-frequency spectrum,
although some studies suggest that the sense of direction also exists at the lowest audible frequencies. In the current
work, a 2AFC localisation experiment is conducted with 18 participants, where the listener must detect a change in
the direction of pure tones and octave band filtered pink noise bursts in the frequency range of 31.5 to 100 Hz. The
angular separations between the low-frequency sound events utilised in the test are 10, 20, 25, 35 and 45 degrees
in the left azimuth plane. The results agree with those studies showing that humans can localise even the lowest
audible frequencies. Changes in direction as small as 10 degrees can be reliably detected for pink noise bursts from
31.5 Hz and for pure tones from 63.5 Hz. The psychoacoustic experiment was conducted in an anechoic room with
minor room resonances which caused a significant change in the directional judgement for the affected frequencies,
demonstrating how strongly room resonances can interact with directional hearing in the low-frequency spectrum.

1 Introduction The perception of low-frequency directionality has

been has been scarcely approached in literature. The
The capacity of humans to detect the location of sound localisation ability of low frequencies has been thought
sources has been actively studied through auditory ex- to be somehow weaker than the localisation of high
periments since the 1900s [1]. However, in unravelling frequencies, especially if the other frequencies are pre-
the mechanisms that enable spatial hearing, one is faced sented to the listener at the same time from another
with a complex task that is strongly dependant upon direction. In literature, there exists at the moment two
the experimental conditions, such as the nature of the major contrasting opinions about low-frequency sound
sounds involved and the attributes of the listening en- source localisation: some argue that it is beyond hu-
vironment. When these attributes vary between the au- mans’ capacity to localise a source below a certain
ditory experiments conducted, disagreements arise be- frequency [2, 3, 4, 5, 6, 7], while others are firmly con-
tween the hypothesis presented by different researchers. vinced that this task is indeed possible down to the
The topic of low-frequency sound source localisation lowest audible frequency [8, 9, 10, 11, 12, 13].
makes no exception. Of the papers which conclude that there is little direc-
Nastasa, Pulkki, Mäkivirta Auditory localisation of low-frequency sound sources

tional information in the low-frequency spectrum, the whether the minute differences in phase between the
experiments have been conducted in small living-room- ears could offer enough directional information.
like listening environments and included conditions The role of time differences in the direction perception
where the position of the sound sources excite a strong of low frequencies has been studied by Braasch et al.
modal behaviour [3, 4, 5]. A number of these authors in [8] and Mohamed and Cabrera in [10]. Braasch et al.
note that the localisation ability might be impeached propose the notion that ITD could be accounted as the
by the presence of room modes, that could negatively main cue used by the auditory system for the left-right
impact the binaural cues presented at the ear of the discrimination, and note further that the presence of
listener. This claim might hold true as the studies that reverberation can distort this cue and create conflicting
support the human ability to localise low-frequencies information. The results of Mohamed and Cabrera
[8, 9, 10, 11, 12, 13] have either been conducted in less confirm that lateralisation based on ITD can occur in
reverberant environments [9, 13], or the placement of very low frequencies, with the strength of the effect
the sound sources did not excite as strong of a modal increasing with frequency. The authors conclude that
behaviour [8]. Such studies [8, 9, 10] have shown that while localisation acuity is poor at very low frequencies,
discrimination between left and right sources is pos- some lateralisation is at least possible at frequencies
sible with high accuracy down to the lowest audible as low as 31.5 Hz (albeit marginal) and 40 Hz, and the
frequencies of human hearing. Moreover, the effect of acuity increases with frequency.
the environment has been also exhibited by Martens
et al. [12] where the discrimination accuracy of sub- 1.2 Room Modes
woofer sources in large low-reverberation spaces was
highly correct, while small reverberant spaces produced Room modes (standing waves) occur as an interaction
inaccurate results. However, no attempt has yet been between the radiating source and its reflections within
made to objectively quantify this localisation ability, the room, when the wavelengths of the source are pro-
nor the extent to which the listening environment has portional to the dimensions of the listening environ-
an influence upon it. ment. These waves result in two types of interference:
constructive interference (maximum pressure nodes)
The current study aims to answer the question whether and destructive interference (minimum pressure nodes).
the auditory system has the ability to detect changes Constructive interference amplifies sound, while de-
in the direction of very low-frequency sound events structive interference reduces sound, resulting in a pat-
(31.5 to 100 Hz) when the variables introduced by the tern of higher and lower sound pressure levels within
listening environment are strongly minimised. For this the room.
purpose, the listening room chosen for this experiment
In the reproduction of the low-frequency spectrum, the
is a large anechoic chamber with a small amount of
sound field is dominated by the presence of modal reso-
modal resonant behaviour. The ability of the auditory
nances which result in uneven energy decay rates across
system to make reliable directional choices about the
the frequency range. These energy decay rates modify
low-frequency spectrum and the spatial resolution in
the magnitude response of the primary sound source
this frequency range are investigated through a psy-
and cause audible artifacts in the form of loudness and
choacoustic experiment.
timbre changes. In terms of auditory localisation, room
modes could potentially hinder the localisation of low
1.1 Spatial hearing in the very low frequency
frequencies as the phase and level of the standing wave
spectrum
are summed up with those of the direct sound at the ear
of the listener. As a result, the ITD and ILD cues are
The main mechanisms of sound source localisation are likely to yield an auditory image unrelated to that of
the interaural time difference (ITD), interaural level dif- the source.
ference (ILD) and the spectral filtering produced by the
pinnae, head and torso (HRTF). In the low-frequency 2 Method
region, as wavelengths are much longer than the size of
the head and the pinna, it is highly unlikely that the ILD This study investigates the spatial resolution of low-
and HRTF are effective cues in providing directional frequency direction perception using a two alterna-
information. In the case of ITDs, it is not yet evident tive forced choice (2AFC) method for pure tones and

AES 154thConvention, Espoo, Helsinki, Finland, 2023 May 13–15

Page 2 of 9
Nastasa, Pulkki, Mäkivirta Auditory localisation of low-frequency sound sources

octave-band-filtered pink noise bursts in the frequency small angular separations between the subwoofers is re-
range of 31.5 to 100 Hz. The spatial resolution was quired. However, as the dimensions of the subwoofers
investigated at the fixed angular separations of 10, 20, are rather large (H 685 x W 718 x D 492 mm), a small
25, 35 and 45 degrees in the left azimuth plane. separation angle between the sound sources cannot be
achieved if they are placed side by side. Therefore, in
2.1 Subjective Listening Test order to achieve a minimum separation angle of 10 de-
grees, one of the subwoofers was stacked on top of the
The ability to perceive the direction of low-frequency two subwoofers placed at 0 and 20 degrees. The sub-
sound sources was measured utilising a relative locali- woofers that were placed on the raised floor net struc-
sation task methodology. Relative localisation studies ture had their drivers pointing downwards, and the one
focus on the spatial resolution of the auditory system on top had its driver pointing upwards. Subwoofers are
by measuring the smallest change in source direction omnidirectional sources, thus the way they are pointed
that the listener can reliably discern. A measure of the should make no difference. Consequently, the test setup
spatial resolution is the minimum audible angle (MAA) configuration had the low-frequency sound sources at
[14], and it can be determined using the psychophysical 0, 10, 20 and 45 degrees in the left azimuth plane.
method of a two-alternative forced-choice (2AFC) test
procedure. In the test, the subject is presented with the Impulse responses (IRs) of each low-frequency sound
first sound, considered the reference position for that source were recorded at the listener’s position using a
trial, followed by the target sound originating from a G.R.A.S. 46AF 1/2 free field microphone. The modal
position to one side of the reference. The subject is behaviour of the anechoic chamber can be observed in
then required to state whether the second auditory event the magnitude response displayed in figure 3. The mag-
lay to the right or the left of the reference. The MAA is nitude responses suggest the presence of one modal
then determined as the angular separation where 75% resonant peak at 50 Hz, and modal resonant notches
of the responses are correct [15]. As this is an introduc- around 31.5 and 80 Hz. Furthermore, the misalign-
tory study in the perception of low-frequency direction, ment in the frequency of the modal peaks and notches
the goal of this experiment is not necessarily to deter- between the different subwoofer positions is visibly
mine the MAA, but more so to utilise the methodology present. It is this misalignment that could potentially
of the MAA procedure in order to find the threshold introduce phase and level differences that are unrelated
where the perception of lateralisation can be considered to those of the actual sound sources and impeach the
reliable. localisation ability. In the listening test design, the fre-
quencies chosen to be studied were 31.5, 40, 50, 63,
2.2 Listening test configuration 80 and 100. Amongst these frequencies, the largest
variation in magnitude response because of modal res-
The listening test was conducted in an anechoic cham- onances were present at 31.5 and 80 Hz, where the
ber that fulfills free field conditions, in accordance with highest level difference between two subwoofers was
ISO 3745 [16] standard, down to 50 Hz, but still shows of 7 dB(SPL), and 5 dB(SPL), respectively.
a slight modal behaviour at low frequencies. The di-
mensions of the room are from a wedge tip to a wedge The listening test setup also included a pair Genelec
tip 8.2 m x 8.2 m (width x length) and the height is 7.2 8020B loudspeakers that played decorrelated pink noise
m, with a raised net floor structure. The measured back- alongside the low-frequency stimuli, with the purpose
ground noise inside the chamber is below the human of masking any high-frequency distortion components
threshold of hearing as defined in ISO 226 [17]. caused by the subwoofers. Such high-frequency com-
ponents are to be avoided as they can offer unwanted
Four Genelec 7380A active subwoofers were used in localisation cues.
the test, with the low frequency cutoff point at 16 Hz
and the high-cutoff at 100 Hz. All subwoofers were 2.3 Stimuli
placed equidistantly from the centre of the listener’s
head. The positioning of the sound sources and the Two types of audio signals were used in the test: pure
listener within the room is displayed in figure 1 and 2. tones and octave-band filtered bursts of pink noise, both
As the goal of this experiment is to determine the hori- generated using Max MSP at a sampling frequency of
zontal spatial resolution in the low frequency spectrum, 48 kHz and 24-bit resolution. Each stimulus had a 100

AES 154thConvention, Espoo, Helsinki, Finland, 2023 May 13–15

Page 3 of 9
Nastasa, Pulkki, Mäkivirta Auditory localisation of low-frequency sound sources

Fig. 2: Listening test subwoofer setup seen from the

perspective of the listener

95

90
Magnitude (dBSPL) 85

80
Fig. 1: Layout of the listening test subwoofer setup
in the anechoic room. Distances to the sound 75
sources refer to the centre of the listener’s head. 70
The angle of each subwoofer in relation to the
listener is indicated in different colours 65
Sub 0° Sub 10° Sub 20° Sub 45°
60
30 40 50 60 70 80 90
ms level ramp-up and ramp-down in order to prevent Frequency (Hz)
any high-frequency clicks.
Fig. 3: Magnitude response of the four subwoofer po-
The stimuli were 1 s long and level-matched for all sitions displayed in dB(SPL)
subwoofers playing at a sound level of 80 dB(C), mea-
sured at the listener position. This playback level was completely mask all the harmonic distortion compo-
chosen as the sound components were clearly audible nents introduced by the subwoofers. An RME Fireface
and comfortable at all frequencies. The pink noise UCX audio interface handled all the audio playing.
bursts were band-passed to be one octave wide, with
the centre frequencies of 31.5, 40, 50, 63, 80 and 100 2.4 Test procedure
Hz. The 31.5 Hz sinusoid was removed from the test
conditions because of the interfering strong room mode The listening test utilised a 2AFC method with the sub-
resonant notch present at this frequency. As it can be ject being forced to choose between either the left or
seen in figure 3, the subwoofers located at 0, 20 and right change in direction of the perceived sound image.
45 degrees are strongly attenuated. Due to this strong The test interface was programmed in Max MSP and
level attenuation, the sinusoids presented by these sub- the presentation of the conditions were randomised. At
woofers could not be reproduced at the required level the start of the test, one of the four subwoofers was
of 80 dB(C) without excessive distortion artifacts that arbitrarily chosen as the reference and the 1 s stimulus
would introduce unwanted localisation cues. was played from this source. After a 100 ms interval,
the same stimulus was presented from another position,
Moreover, to avoid the remaining minimal high- either to the left or to the right of the reference. The
frequency distortion components introduced by the subject was then asked: In which direction is the sound
subwoofers, especially audible when playing pure moving? Press either the left or right arrow to mark
tones, decorrelated full-spectrum pink noise was played your answer. By only having to press the arrows on the
through the two separate loudspeakers at a sound level keyboard, the interface was intended to be as unobtru-
of 58 dB(C). This level was found to be sufficient to sive as possible so that the subject could concentrate on

AES 154thConvention, Espoo, Helsinki, Finland, 2023 May 13–15

Page 4 of 9
Nastasa, Pulkki, Mäkivirta Auditory localisation of low-frequency sound sources

Separation Subwoofer 1 Subwoofer 2 2.5 Subjects

Angle Direction Direction
(deg) (deg) (deg) In total, 18 subjects took part in the test, with an av-
10 0 10 erage age of around 30. All subjects reported normal
10* 10 20 hearing. The participants were either researchers at
20 0 20 the Acoustics Lab of Aalto University, students of the
25 20 45 Acoustics and Audio Technology major or people with
35 10 45 a musical background.
45 0 45
The subjects were instructed to face the centre of the
Table 1: All the separation angles used in the test and subwoofer located at 0 degrees and no head movements
the subwoofer positions to form the separa- were allowed during the stimuli presentation. The chair
tion angles of the participant was placed on a wooden structure in
order to minimise the vibration of the wire net floor
structure.
the task with either eyes open or closed. Before indi- The participants were also advised to take breaks dur-
cating their response, the subject was forced to wait for ing the experiment if listening fatigue started to occur.
the whole duration of the stimuli presentation. This en- Most subjects had previous experience with localisa-
sured that all participants listened equally to the whole tion listening tests. In order to insure that everybody
length of the stimuli. After indicating their response, is familiar with the task, the testing method and the
the test would progress to the next trial, where a new interface, a training session was provided for all partic-
reference position would be chosen. ipants.

With the four subwoofer positions, five angular separa- 3 Results

tions can be formed for the test, namely: 10, 20, 25, 35
and 45 degrees. The subwoofer positions which form The average results across all participants, categorised
each angle separation utilised in the test are displayed on stimulus type as a function of the angular separation
in Table 1. are displayed in figure 4 and 5, where the dotted black
line denotes the threshold of MAA and the solid line
Although the angular separation between the sub- represents the 50% threshold of chance. In order to
woofers located at 0 and 10 degrees is the same as display the inter-subject variability, the results of the
the one formed by the subwoofers located at 10 and 20 psychoacoustic experiment are also displayed in the
degrees, it is expected that the localisation performance form of box-plots in figure 6 for both the pink noise
will be different for these two cases. The latter sep- stimuli and the pure tones, for each frequency as a
aration represents the case of localisation away from function of the angular separation. The central red
the median plane, and research [14, 15, 18] shows that mark indicates the median, the top and bottom edges
the localisation abilities decrease when the reference are the 25th and 75th percentiles, and the outliers are
azimuth moves away from the median plane. For this plotted using the ’+’ symbol. The angle separation of
reason, these two angular separations were treated as 10 degrees marked with an asterisk indicates the case
different cases. of localisation away from the median plane, whereas
the 10 degrees case represents the angular separation
With 6 angular separations, 2 changes in direction (left formed by the subwoofers located at 0 and 10 degrees.
or right), 6 frequencies for the pink noise, 5 frequencies
of the pure tones and 5 repetitions of each condition, the Overall, the results show that the subjects have the abil-
test resulted in a total number of 660 trials. In average, ity to detect the direction of the low-frequency sound
the subjects took about 27 minutes to complete the test. events, and this ability is a function of the stimulus type,
frequency and angle of separation. In the case of the
pink noise stimulus, the localisation ability is clearly
demonstrated in all frequency bands. The results dis-
play an increase in the percentage of correct responses

AES 154thConvention, Espoo, Helsinki, Finland, 2023 May 13–15

Page 5 of 9
Nastasa, Pulkki, Mäkivirta Auditory localisation of low-frequency sound sources

Pink Noise 31.5 Hz Pink Noise 40 Hz Sinusoid 40 Hz Sinusoid 50 Hz

100 100 100 100
Correct Responses (%)

Correct Responses (%)

Correct Responses (%)

Correct Responses (%)

80 80
80 80
60 60

60 60 40 40

20 20
40 40
°*

°*

°*

°*
°
°
°

°
°
°

°
°
°

°
°
°

°
10
20
25

35

45

10
20
25

35

45

10
20
25

35

45

10
20
25

35

45
10

10

10

10
Angular Separation Angular Separation Angular Separation Angular Separation
Pink Noise 50 Hz Pink Noise 63 Hz Sinusoid 63 Hz Sinusoid 80 Hz
100 100 100 100
Correct Responses (%)

Correct Responses (%)

Correct Responses (%)

Correct Responses (%)

80 80
80 80
60 60

60 60 40 40

20 20
40 40
*

*
°
°
°

°
°
°

°
°
°

°
°
°

°
°

°
10
20
25

35

45

10
20
25

35

45

10
20
25

35

45

10
20
25

35

45
10

10

10

10
Angular Separation Angular Separation Angular Separation Angular Separation
Pink Noise 80 Hz Pink Noise 100 Hz Sinusoid 100 Hz
100 100 100
Correct Responses (%)

Correct Responses (%)

Correct Responses (%)

80
80 80
60

60 60 40

20
40 40
*

*
°
°
°

°
°
°

°
°
°

°
°

°
10
20
25

35

45

10
20
25

35

45

10
20
25

35

45
10

10

10

Angular Separation Angular Separation Angular Separation

Fig. 4: Mean percentage of correct responses across Fig. 5: Mean percentage of correct responses across
all participants for pink noise as a function of all participants for pure tones as a function of
angular separation, categorised on 6 different angular separation, categorised on 6 different
frequencies. The angle separation of 10 degrees frequencies. The angle separation of 10 degrees
marked with an asterisk indicates the case of marked with an asterisk indicates the case of
localisation away from the median plane. localisation away from the median plane.

In fact, the only percentage of correct responses under

with increasing the angular separation, whereas the the threshold of MAA for the pink noise stimulus has
localisation ability also increases with increasing the been recorded at the angular separation of 10 degrees
frequency. For all frequency bands of the pink noise away from the median plane for the frequencies under
stimuli, a separation angle between 20 and 45 degrees 63 Hz. For the frequencies of 63 Hz and higher, the
resulted in over 80% of correct responses. The poorest percentage of correct responses was above 80 for all
localisation performance for this stimulus type is of angular separations.
60% at the angular separation of 10 degrees away from
the median plane. The case of 10 degrees away from Overall, the pink noise had a better localisation per-
the median plane had consistently a worse performance formance than the pure sinusoids, which is expected.
than the angular separation located in the median plane. The reason for this is that the pink noise bursts have a
This result is in accordance to the findings of Mills larger bandwidth and provide more transient informa-
[14], in that the localisation abilities decrease when the tion [18], while also being less susceptible to modal
reference azimuth moves away from the median plane. effects because of the broader frequency content. In

AES 154thConvention, Espoo, Helsinki, Finland, 2023 May 13–15

Page 6 of 9
Nastasa, Pulkki, Mäkivirta Auditory localisation of low-frequency sound sources

the case of the pure tones of frequencies under 63 Hz, angular separations have been strongly affected, as the
the changes in angular separation had little effect on phase and level of the standing wave modify the charac-
the percentage of the correct responses. For this fre- teristics of the direct sound at the ear of the listener. As
quency range, the percentage of correct responses did a result, for these angular separations, the ITD and ILD
not go above the threshold of MAA, signifying that a cues likely yielded an auditory image unrelated to that
reliable change in direction could not be perceived. For of the source, thus negatively impacting the localisation
frequencies of 63 Hz and over, the localisation ability ability in this frequency band. This effect could also be
increased significantly and the lowest percentages were affirmed by the fact that the sinusoids of 63 and 100 Hz
recorded for the separation angle of 10 degrees away had a strong overall localisation performance and an
from the median plane, with the exception of 80 Hz fre- even magnitude across all subwoofer positions, which
quency. The case of the 80 Hz sine tone band displayed indicates the absence of room modal resonances.
results that do not follow the expected trend set by the
The phenomena observed at the 80 Hz pure tone of this
frequency of 63 and 100 Hz. The angular separation
psychoacoustic experiment have motivated the authors
of 10, 20 and 25 degrees were under the threshold of
to study the effect of room modal resonances on the
the MAA, whilst the angular separation of 10 degrees
perception of low-frequency sound sources in [19]. The
away from the median plane recorded a percentage of
study has demonstrated that room modal resonances,
correct responses just above this threshold. For the
especially the minimum pressure nodes, have a strong
sinusoid of 100 Hz, a reliable change in direction could
effect on the localisation of very low-frequency sound
be perceived regardless of the angular separation.
sources and are the cause for the little directional infor-
mation exhibited in this frequency range.
4 Discussion
5 Conclusion
The results of the psychoacoustic experiment demon-
strate the ability of the auditory system to detect In this study, the extent of the human ability to per-
changes in the direction of low-frequency sound ceive changes in the direction of low-frequency sound
sources in an anechoic environment. This ability is sources was studied through a psychoacoustic experi-
highly dependent on the stimulus type, frequency and ment in an anechoic chamber. The results of the 2AFC
the angular separation between the sound sources. In localisation experiment, conducted with 18 participants,
the case of the pink noise stimulus, the listeners could agree with the studies showing that humans can localise
reliably detect changes in direction for all frequencies even the lowest audible frequencies, and indicate that
in the range of 31.5 to 100 Hz for angular separations the spatial resolution of the auditory system in the low-
ranging from 10 to 45 degrees. frequency spectrum goes beyond a left-right discrimi-
nation. Changes in direction as small as 10 degrees can
The angular separation had a stronger impact on the
be reliably detected for pink noise bursts from 31.5 Hz
localisation ability of pure tones, which were more dif-
and for sinusoids from 63.5 Hz. In addition, the locali-
ficult to localise. For frequencies under 63 Hz, an angu-
sation ability has been greatly impeded at frequencies
lar separation higher than 45 degrees would be needed
which present strong modal resonances, demonstrat-
for a reliable discern of changes in direction. The case
ing that the directional hearing of low-frequency sound
of the 80 Hz pure tone exhibited unusual responses,
sources highly depends on the acoustical properties of
where the localisation performance was strongly im-
the listening environment.
peached for the angular separations of 10, 20 and 25
degrees. As it has been already noted in section 2.2, Nevertheless, the results of this research suggest that
figure 3, this frequency region exhibits the presence the ability to localise low-frequency sound sources is
of strong room modal resonances. The subwoofers lo- an innate quality of the human auditory system.
cated at 0 and 20 degrees are in a modal resonant notch,
whilst the two other subwoofers are unaffected. All 6 Acknowledgements
the angular separations that had percentages under the
MAA threshold at this frequency were formed with ei- The authors would like to extend their gratitude to all of
ther one of the subwoofers located in this modal notch. the researchers and professors of the Aalto Acoustics
This indicates that the perception of direction for these Lab for their vast knowledge and warm company. We

AES 154thConvention, Espoo, Helsinki, Finland, 2023 May 13–15

Page 7 of 9
Nastasa, Pulkki, Mäkivirta Auditory localisation of low-frequency sound sources

Pink Noise 31.5 Hz Pink Noise 40 Hz Sinusoid 40 Hz Sinusoid 50 Hz

100 100 100
100

Correct Responses (%)

Correct Responses (%)

Correct Responses (%)

Correct Responses (%)

80 80
80 80
60 60

60 60 40 40

20 20
40 40
0 0
°*

°*

°*
°
°
°
°
°

°
°
°
°
°

°
°
°
°
°
°*
°
°
°
°
°

10
20
25
35
45

10
20
25
35
45

10
20
25
35
45
10
20
25
35
45

10

10

10
10

Angular Separation Angular Separation Angular Separation Angular Separation

Pink Noise 50 Hz Pink Noise 63 Hz Sinusoid 63 Hz Sinusoid 80 Hz
100 100 100 100
Correct Responses (%)

Correct Responses (%)

Correct Responses (%)

Correct Responses (%)

80 80
80 80
60 60

60 60 40 40

20 20
40 40
0 0
°*

°*

°*

°*
°
°
°
°
°

°
°
°
°
°

°
°
°
°
°

°
°
°
°
°
10
20
25
35
45

10
20
25
35
45

10
20
25
35
45

10
20
25
35
45
10

10

10

10

Angular Separation Angular Separation Angular Separation Angular Separation

Pink Noise 80 Hz Pink Noise 100 Hz Sinusoid 100 Hz
100 100 100
Correct Responses (%)

Correct Responses (%)

Correct Responses (%)

80
80 80
60

60 60 40

20
40 40
0
°*

°*
°*
°
°
°
°
°

°
°
°
°
°
°
°
°
°
°
10
20
25
35
45

10
20
25
35
45
10
20
25
35
45
10

10
10

Angular Separation Angular Separation Angular Separation

Fig. 6: Boxplots of the percentages of correct responses for all listeners grouped on 2 stimuli type, 6 frequencies
and 6 angular separations. The angle separation of 10 degrees marked with an asterisk indicates the case of
localisation away from the median plane

AES 154thConvention, Espoo, Helsinki, Finland, 2023 May 13–15

Page 8 of 9
Nastasa, Pulkki, Mäkivirta Auditory localisation of low-frequency sound sources

are also profoundly grateful to the participants of the [9] Subkey, A., Cabrera, D., and Ferguson, S., “Local-
listening test, who so willingly donated their time to ization and image size effects for low frequency
listen to bass. sound,” in 118th Convention of the AES, paper,
volume 6325, 2005.
References [10] Mohamed, M. and Cabrera, D., “Human sensi-
tivity to interaural phase difference for very low
[1] Strutt, J. W., “On our perception of sound direc- frequency sound,” in Acoustics 2008-Australian
tion,” Philosophical Magazine, 13(74), pp. 214– Acoustical Society Conference, p. 47, 2008.
32, 1907. [11] Hill, A. J., Lewis, S. P., and Hawksford, M. O.,
“Towards a generalized theory of low-frequency
[2] Borenius, J., “Perceptibility of direction and time
sound source localization,” 2012.
delay errors in subwoofer reproduction,” in Au-
dio Engineering Society Convention 79, Audio [12] Martens, W. L., Braasch, J., and Ryan, T. J., “Spa-
Engineering Society, 1985. tial auditory display using multiple subwoofers
in two different reverberant reproduction environ-
[3] Theile, G. and Kügler, C., “Loudspeaker repro- ments,” in ICAD. Eleventh meeting of the interna-
duction: Study on the subwoofer concept,” in tional conference on auditory display, 2005.
Audio Engineering Society Convention 92, Audio
Engineering Society, 1992. [13] Martens, W. L., “Subjective evaluation of auditory
spatial imagery associated with decorrelated sub-
[4] Welti, T. S., “Subjective comparison of single woofer signals,” Georgia Institute of Technology,
channel versus two channel subwoofer reproduc- 2002.
tion,” in Audio Engineering Society Convention
117, Audio Engineering Society, 2004. [14] Mills, A. W., “On the minimum audible angle,”
The Journal of the Acoustical Society of America,
[5] Zacharov, N., Bech, S., and Meares, D. J., “The 30(4), pp. 237–246, 1958.
Use of Subwoofers in the Context of Surround
Sound Program Reproduction,” in Audio Engi- [15] Blauert, J., Spatial hearing: the psychophysics of
neering Society Convention 102, Audio Engineer- human sound localization, MIT press, 1997.
ing Society, 1997. [16] ISO 3745:2012, “Acoustics. Determination of
Sound Power Levels of Noise Sources using
[6] Kelloniemi, A., Ahonen, J., Paajanen, O., and Sound Pressure Precision Methods for Anechoic
Pulkki, V., “Detection of subwoofer depending and Hemi-Anechoic Rooms,” Standard, Interna-
on crossover frequency and spatial angle between tional Organization for Standardization, Geneva,
subwoofer and main speaker,” in 118th Conven- CH, 2012.
tion of the AES, paper, volume 6431, Citeseer,
2005. [17] ISO 226:2003, “Acoustics — Normal equal-
loudness-level contours.” Standard, International
[7] Rämö, J., Bergen, S., Parker, J., Yli-Kätkä, V.- Organization for Standardization, Geneva, CH,
M., and Pulkki, V., “Detection of Two Sub- 2003.
woofers: Effect of Broad-Band-Channel Level
and Crossover Frequency,” in Audio Engineer- [18] Moore, B. C., An introduction to the psychology
ing Society Convention 132, Audio Engineering of hearing, Brill, 2012.
Society, 2012.
[19] Nastasa, M., Pulkki, V., and Mäkivirta, A., “Im-
[8] Braasch, J., Martens, W. L., and Woszczyk, W., pact of standing waves on human auditory per-
“Modeling auditory localization of subwoofer sig- ception of low-frequency direction,” in Audio En-
nals in multi-channel loudspeaker arrays,” in Au- gineering Society International Conference on
dio Engineering Society Convention 117, Audio Spatial and Immersive Sound, Audio Engineering
Engineering Society, 2004. Society, 2023.

AES 154thConvention, Espoo, Helsinki, Finland, 2023 May 13–15

Page 9 of 9