"Listening through different ears in the Sydney Opera House"


Return to CARLab Research Page.

 

SOH recording setup

Binaural head-related impulse response (HRIR) measurements were conducted in the Concert Hall of the Sydney Opera House using a Soundsphere loudspeaker and a customized sub-frequency loudspeaker at 6 x positions on the stage platform marked in letter A, B, C, D, E, F. Measurements were made in 3 seats (N35, P34,Q33) for 7 subjects and 1 acoustic mannequin which was a customized model (ears, head, and torso) see Figure 1 below.

 

 

Figure 1: a) Overview of recording setup – 6 loudspeaker positions on the stage platform marked A to F, measured 3 seating: N35 marked in orange circle, P34 marked in red circle and Q33 marked in green.

 

Overview

  • Subjects were instructed to directly face the center stage by pointing their nose to the sound source and sit squarely and aligned with the back of their chair. However, they were not physically constrained.
  • The HRIR measurements were recorded on an Alesis HD24 Hard Disk Recorder using seven log sine sweeps from 20 Hz to 22 kHz with a duration of approximately five seconds. Subjects were instructed to directly face the center stage and sit squarely and aligned with the back of their chair.
  • After the binaural HRIRs were recorded, they were windowed to 2.6 seconds in length and their associated interaural time difference (ITD) value and interaural level difference (ILD) value were analyzed through listening tests and calculations using the MATLAB software.

1. Binaural HRIR

Here, we will present VAS sound simulation for binaural HRIR collected in seat P34 from center stage loudspeaker marked "F".

 

2 subjects' HRIRs were removed from the data set, leaving a total of 6 binaural HRIRs. Beacuse the primary issue is that a subject may not have been directly facing the centre stage leading to an ‘artificial' ILD and ITD cue than can be used to discriminate between different binaural HRIRs.

 

 


Subject
HRIR (click links below to download)
1
2
3
4
5
6

Table 1: 6 subjects' recorded binaural HRIR in seat P34.

 

2. Sound Stimuli

A variety of sound stimuli were chosen ranging from speech, instrumental music, ensemble music, and orchestral music (see Table below). The sound excerpts covered different frequency ranges and textural complexities and were limited to 10-15 seconds in length.

 

 

Anechoic Sound Source
Abbreviated Song Name
(click link below to download)
Track No
Music for Archimedes CD Guitar 15, “Etude No.6 in E minor” by H.Villa Lobos
Cello 20,”Theme” by Weber
Xylophone 27,” Sabre Dance” by Khachaturian
Female voice 4, Female speech
Denon Professional Test CD 2 Flute 31
Jazz ensemble 35
Orchestra - Mozart 25, “ Die Hochzeit Von Fiagro” by Mozart
Orchestra - Shostakovich 26, “Symphony No.5” by Shostakovich
White noise Random generated broadband noise

Table 2: 9 mono anechoic excerpts are extracted from two CD archives, ranging from speech, tonal instruments to complex ensemble and orchestral pieces.

 

3. Binaural VAS stimuli

A total of 54 VAS sound stimuli were generated by convolving the binaural HRIR filters with the 9 mono anechoic sound stimuli described above.

 


No.
Mono Sound Source
Subject

Convolved VAS Sounds for P34

(click links below to download)

1
Guitar
A
2
G
3
B
4
R
5
J
6
H
7
Cello
A
8
G
9
B
10
R
11
J
12
H
13
Xylophone
A
14
G
15
B
16
R
17
J
18
H
19
Female voice
A
20
G
21
B
22
R
23
J
24
H
25
Flute
A
26
G
27
B
28
R
29
J
30
H
31
Jazz ensemble
A
32
G
33
B
34
R
35
J
36
H
37
Orchestra - Mozart
A
38
G
39
B
40
R
41
J
42
H
43
Orchestra - Shostakovich
A
44
G
45
B
46
R
47
J
48
H
49
White noise
A
50
G
51
B
52
R
53
J
54
H

 

Table 3: 54 binaural VAS sounds generated by convolving 9 mono anechoic sounds with 6 subjects' recorded HRIRs for seat P34.

 

 

NOTE:

 

The correctness of the processing steps described above was verified by comparing the simulated binaural VAS sound stimuli for the Orchestra-Moz. piece with a true binaural recording of the same piece recorded in the same session as when the binaural HRIRs were recorded. This page was produced by Angela Li.