close

Вход

Забыли?

вход по аккаунту

?

2017 AJA-16-0101

код для вставкиСкачать
AJA
Research Article
Combined Effects of Noise and Reverberation on
Sound Localization for Listeners With Normal
Hearing and Bilateral Cochlear Implants
Yunfang Zheng,a Janet Koehnke,b and Joan Besingb
Purpose: This study examined the individual and combined
effects of noise and reverberation on the ability of listeners
with normal hearing (NH) and with bilateral cochlear
implants (BCIs) to localize speech.
Method: Six adults with BCIs and 10 with NH participated.
All subjects completed a virtual localization test in quiet
and at 0-, −4-, and −8-dB signal-to-noise ratios (SNRs) in
simulated anechoic and reverberant (0.2-, 0.6-, and 0.9-s
RT60) environments. BCI users were also tested at +8- and
+4-dB SNR. A 3-word phrase was presented at 70 dB SPL
from 9 simulated locations in the frontal horizontal plane
(±90°), with the noise source at 0°.
Results: BCIs users had significantly poorer localization than
listeners with NH in all conditions. BCI users’ performance
started to decrease at a higher SNR (+4 dB) and shorter RT60
(0.2 s) than listeners with NH (−4 dB and 0.6 s). The combination
of noise and reverberation began to degrade localization of BCI
users at a higher SNR and a shorter RT60 than listeners with NH.
Conclusion: The clear effect of noise and reverberation
on the performance of BCI users provides information that
should be useful for refining cochlear implant processing
strategies and developing cochlear implant rehabilitation
plans to optimize binaural benefit for BCI users in everyday
listening situations.
T
1999a). In addition, reflected sounds from reverberation
may also create masking by smearing energy across frequencies and, depending on the reverberation time (RT), degrade
the original sound affecting the ILDs and interrupting the
ITDs, increasing difficulty in localizing sounds (Hartmann,
1983). A cochlear implant (CI) is an electronic device that
can help restore hearing for patients with severe-to-profound
sensorineural hearing loss. Bilateral implantation does improve spatial hearing in noise and reverberation (Grantham,
Ashmead, Ricketts, Labadie, & Haynes, 2007; Neuman,
Haravon, Sislian, & Waltzman, 2007; Schön, Müller, Helms,
& Nopp, 2005; Van Hoesel & Tyler, 2003; Verschuur, Lutman,
Ramsden, Greenham, & O’Driscoll, 2005; Zheng, Koehnke,
& Besing, 2016). However, noise and reverberation still
present a problem for CI users. Due to limitations of current speech processing strategies and compression circuits,
interaural cues may not be fully preserved (Preece, 2010).
As indicated in our preliminary study (Zheng, Koehnke,
Besing, & Spitzer, 2011), listeners with bilateral CIs (BCIs)
had significantly poorer localization performance in noise
and/or reverberation. That original study included two BCI
users who were tested in one reverberant environment. Thus,
it is not clear how noise plus reverberation systematically
affects listeners with NH and with BCIs. This study was designed to investigate the ability of these groups of listeners
to localize speech in noise and reverberation.
he ability to localize sound relies primarily on binaural processing, that is, sensitivity to interaural
level difference (ILD) and interaural time difference
(ITD). It is important for both communication and safety.
Typical listening environments contain noise and reverberation, which degrade binaural cues and result in decreased
sound localization ability for listeners with normal hearing (NH) and impaired hearing (e.g., Hartmann, 1983;
Lorenzi, Gatehouse, & Lever, 1999a, 1999b). Background
noise may mask the signal and impact both binaural cues
and monaural spectral cues, especially the ITDs; as a result,
the ILDs and the spectral cues are the dominant cues for
localization in noise (Lorenzi et al., 1999a). ITDs are the
predominant cue for low-frequency sounds, and because
the head shadow effect is not present at low frequencies,
ITDs contribute less to localization in noise compared with
ILDs and spectral cues (e.g., Gelfand, 2009; Lorenzi et al.,
a
Department of Communication Sciences and Disorders, Central
Michigan University, Mount Pleasant, MI
b
Department of Communication Sciences and Disorders, Montclair
State University, Bloomfield, NJ
Correspondence to Yunfang Zheng: zheng4y@cmich.edu
Editor: Sumitrajit (Sumit) Dhar
Associate Editor: Ryan McCreery
Received October 12, 2016
Revision received March 1, 2017
Accepted July 12, 2017
https://doi.org/10.1044/2017_AJA-16-0101
Disclosure: The authors have declared that no competing interests existed at the time
of publication.
American Journal of Audiology • 1–12 • Copyright © 2017 American Speech-Language-Hearing Association
Downloaded From: http://aja.pubs.asha.org/ by a Fudan University User on 10/28/2017
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
1
Studies of listeners with NH have revealed that localization accuracy declines with decreasing signal-to-noise
ratio (SNR) and localization performance starts to be affected
when the SNR in the frontal horizontal plane goes below
about 0–4 dB (e.g., Good & Gilkey, 1996; Kerber & Seeber,
2012; Lorenzi et al., 1999a). Lorenzi et al. also found that
localization performance is affected by the position of the
noise source. When the noise was at ±90°, localization was
poorer and more variable than when the noise was at 0°
azimuth, for SNRs below +6 dB. This is because, compared
with noise presented at 0°, noise presented from other locations will have a greater impact on interaural cues, especially
on the ILDs, increasing localization difficulty. Studies of
listeners with NH have shown that localization performance
is consistently poorer in a reverberant room than in an
absorbent room (Giguere & Abel, 1993) and accuracy of
localization of continuous broadband noise decreases significantly with increasing RT (Hartmann, 1983; Rychtarikova,
Van den Bogaert, Vermeir, & Wouters, 2007). In a virtual
localization study by Besing and Koehnke (1995), 15 listeners with NH (five adults and 10 children [7–12 years old])
localized speech in anechoic (AN) and reverberant (RT60 =
0.2 s) environments. The results showed comparable localization performance for both adults and children with NH,
consistent with what was reported in the literature based
on sound-field testing. This indicates that their virtual test
is a feasible test for use with children and adults with NH.
However, Besing and Koehnke’s results did not show a reverberation effect; this might be because of the shorter RT used
compared with the other studies described above.
Only a few studies have investigated the combined
effects of noise and reverberation on sound localization in
listeners with NH. Kopco, Best, and Shinn-Cunningham
(2007) recruited seven subjects with NH to localize a single
click at 0-dB SNR in both AN and reverberant rooms
(RT = 0.4–0.6 s). The results showed greater response variability and larger localization errors in the reverberant
room than in the AN room for the lateral noise sources,
suggesting a possible adverse effect of combined noise
and reverberation. However, the combined effects were
not observed for listeners with NH in Zheng et al. (2011).
Seven listeners with NH localized speech in quiet and 0- to
−8-dB SNR in one reverberant environment (RT60 = 0.2 s).
It is likely that a combined effect was not observed in the
2011 study due to the short RT and the fact that no lateral
noise sources were used.
A few studies have investigated BCI users’ localization
performance in noise. Van Hoesel et al. (2008) recruited four
BCI users to locate a click train (100 and 500 Hz) at 0-dB
SNR and determined that BCI users had relatively poorer
localization performance when the noise was located at 90°
azimuth than at 0° azimuth. Mosnier, Sterkers, and Bebear
(2009) found a binaural advantage in noise (SNR of 15 dB)
for BCI users. Kerber and Seeber (2012) recruited 10 BCI
users; they listened to noise pulses with background noise at
SNRs from 0 to 10 dB. Their results showed that BCI users
have poorer localization performance than listeners with NH
and their localization accuracy decreases as SNR decreases.
2
There have been two studies that investigated the
effect of reverberation on BCI users’ localization performance. Kerber and Seeber (2013) compared BCI users’
localization performance in a simulated reverberant room,
RT60 = 0.4 s, with that in an AN room. The results showed
a significant effect of reverberation on BCI users but not
on listeners with NH. Zheng et al. (2016) recruited six BCI
users to listen to a three-word phrase in simulated reverberant
environments at RT60s from 0 to 0.9 s in quiet. The results
indicated that reverberation has more adverse effects on
localization performance for listeners with BCIs than those
with NH, and BCI users’ localization accuracy decreases
beginning at a shorter RT60 (0.6 s) than that of listeners with
NH (0.9 s). These data are included in this article for comparison purposes in the Discussion and Conclusion section.
However, other than our preliminary study (Zheng
et al., 2011), no study has systematically investigated the
effects of noise and the combined effects of noise plus reverberation on localization performance for BCI users. As sound
localization is crucial for both communication and safety, it
is important to understand how BCI users locate sounds in
realistic listening situations. The virtual localization test developed by Besing and Koehnke (1995) overcomes many of
the problems of free-field localization tests (e.g., the ability to
replicate source locations in actual rooms, head movement),
although generic head-related transfer functions (HRTFs)
may not fully represent the real HRTFs of an individual.
It is a straightforward way to assess performance and a sensitive tool for listeners with NH and impaired hearing. In
our preliminary study (Zheng et al., 2011), seven listeners
with NH and two BCI users listened to a three-word phrase
under various SNRs in both simulated AN and reverberant
environments (RT60 = 0.2 s). The results showed a significant
combined effect of noise and reverberation on BCI users
and further confirmed the feasibility of using the virtual
localization test (Besing & Koehnke, 1995) to assess performance of listeners with NH and BCIs. This study was designed to investigate how noise and reverberation affect
localization ability of listeners with NH and BCIs for a
wider range of RTs and with more subjects. A virtual localization procedure was used (a) to determine how localization performance is affected by changing SNR and RT and
(b) to compare listeners with NH and BCIs in regard to performance in noise and/or reverberant environments.
Method
Subjects
Ten adults with NH (age = 49–64 years, mean =
54.5 years) and six adults with postlingual deafness and
BCIs (age = 49–69 years, mean = 58.3 years) were recruited.
Background information for the subjects with BCI is provided in Table 1. All of the subjects were native English
speakers, and none of them had a history of neurological
pathology. The detailed information concerning the CI
devices and pre- and postimplant audiologic test results
for the subjects with BCI is provided in Tables 2 and 3,
American Journal of Audiology • 1–12
Downloaded From: http://aja.pubs.asha.org/ by a Fudan University User on 10/28/2017
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
Table 1. Background information for subjects with bilateral cochlear implant (BCI).
Age and
gender
CI device
and processor
1
62 years, F
2
65 years, M
3
52 years, F
4
69 years, F
5
49 years, F
6
53 years, M
ABC
HiRes 90 k
Harmony
L+R
ABC
HiRes 90 k
Harmony
L+R
CC
CI24RE (CA)
Freedom
L+R
CC
CI24RE (CA)
Freedom
L+R
CC
L: CI24RE (CA)
R: CI24R (CS)
Freedom
CC
CI24RE (CA)
Freedom
L+R
Subject
Zheng et al.: Noise and Reverberation on BCI Sound Localization
CI strategy
Age at
implantation
BCI
experience
Age received
hearing aids
Hearing aid
experience
Etiology and onset
of hearing loss
HiRes 120
L+R
L: 49 years
R: 59 years
26 months
L: 24 years
R: 24 years
L: 8 years
R: 25 years
Unknown
Age = 20 years
HiRes 120
L+R
L: 64 years
R: 53 years
24 months
L: 17 years
R: 18 years
L: 36 years
R: 26 years
Unknown
Age = 10 years
ACE
L+R
L: 50 years
R: 51 years
18 months
L: 46 years
R: 46 years
L: 4 years
R: 5 years
Unknown
Age = 31 years
ACE
L+R
L: 67 years
R: 67 years
31 months
N/A
N/A
Bacterial meningitis
Age = 67 years
ACE
L+R
L: 46 years
R: 38 years
27 months
L: 15 years
R: 15 years
L: 31 years
R: 23 years
Unknown
Age = 15 years
L: ACE
R: CIS
L: 52 years
R: 51 years
14 months
L: 47 years
R: 47 years
L: 4 years
R: 4 years
Meniere’s disease
Age = 42 years
Note. ABC = Advanced Bionics Corporation; ACE = advanced combinational encoder; CA = contour advance; CC = Cochlear Corporation; CI = cochlear implant; CIS = continuous
interleaved sampling; F = female; HiRes = high resolution; L = left ear; M = male; R = right ear.
3
Downloaded From: http://aja.pubs.asha.org/ by a Fudan University User on 10/28/2017
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
Table 2. Detailed information about cochlear implant devices of subjects with bilateral cochlear implant (BCI).
Subject
Device
Manufacturer
Model
Microphone location
Programming
parameters
Strategy
Stimulation mode
Pulse rate (pulses/s)
Pulse width (μs)
Dynamic range (dB)
L
R
L
R
L
R
L
R
L
R
L
R
L
R
L
R
S1
S2
S3
S4
S5
S6
ABC
ABC
HiRes 90 k
HiRes 90 k
Canal entrance
Canal entrance
HiRes 120
HiRes 120
MP
MP
55,680
41,250
18
24.2
60
60
ABC
ABC
HiRes 90 k
HiRes 90 k
Canal entrance
Canal entrance
HiRes 120
HiRes 120
MP
MP
55,680
55,680
18
18
65
60
CC
CC
CI24RE (CA)
CI24RE (CA)
Upper helix
Upper helix
ACE
ACE
MP1 + 2
MP1 + 2
7,200
7,200
25
25
47–75
41–60
CC
CC
CI24RE (CA)
CI24RE (CA)
Upper helix
Upper helix
ACE
ACE
MP1 + 2
MP1 + 2
7,200
7,200
25
25
47
41–52
CC
CC
CI24RE (CA)
CI24R (CS)
Upper helix
Upper helix
ACE
ACE
MP1 + 2
MP1 + 2
8,640
8,640
25
25
46
23
CC
CC
CI24RE (CA)
CI24RE (CA)
Upper helix
Upper helix
ACE
CIS
MP1 + 2
MP1
7,200
10,800
25
25
47–72
109–124
Note. None of the cochlear implant devices used by the subjects in this study had a directionality feature. ABC = Advanced Bionics
Corporation; ACE = advanced combinational encoder; CA = contour advance; CC = Cochlear Corporation; CIS = continuous interleaved
sampling; HiRes = high resolution; L = left ear; MP = monopolar; R = right ear; S = subject.
respectively. Overall, hearing sensitivity of subjects with
BCI was within normal limits (25 dB HL or better) or in
the normal-to-mild hearing loss range (25–35 dB HL) with
their CIs, and all of them had excellent speech understanding
ability in quiet. NH was defined as having NH sensitivity
(pure-tone thresholds of 25 dB HL or better at octave frequencies from 250 to 8000 Hz) and normal middle ear
function (tympanogram with normal peak pressure, compliance, and ear canal volume). The use of human subjects
for this project was approved by the institutional review
board, and each subject signed a consent form before beginning the experiment.
Stimuli and Listening Conditions
The target stimulus/signal was a three-word phrase,
“mark the spot.” The masker was a speech-spectrum noise,
which was randomly selected from 20 samples for each
trial; each noise had the same duration as the target stimulus
and was presented simultaneously with the signal. The noise
was processed to simulate a source at 0° azimuth, and its
level was varied to achieve different SNRs.
The signal was processed to simulate nine source
locations in the frontal horizontal plane (±90°), 22.5° apart.
At 0°, the signal level was set at 70 dB SPL, and the level
at each ear for sources at other locations varied due to the
head shadow effect.
Localization ability was assessed at four SNRs (quiet
and 0-, −4-, and −8-dB SNR) and four RT60s (0, 0.2, 0.6,
and 0.9 s). BCI users were also tested at +8- and +4-dB SNR.
Signal Processing
The virtual localization test developed by Besing and
Koehnke (1995) was used in this study. To simulate the
sound-field conditions, HRTFs were obtained for different
room conditions (AN and reverberant) for each ear and
each sound location (nine locations in the frontal horizontal
plane from −90° to +90° in 22.5° steps). Then, the HRTFs
were used to convolve the signal and the noise separately.
Finally, the noise and signal for each ear were combined
Table 3. Preimplantation and postimplantation sound-field testing results for subjects with bilateral cochlear implant (BCI).
Preimplant (% correct) aided
Subject
S1
S2
S3
S4
S5
Postimplant (% correct) with BCIs
S6
Avg.
CNC words
0
0
8
0
12
12
5
CNC phonemes
0
0
22
0
33
38
16
HINT in quiet
0
0
28
0
36
61
21
Warble tones Moderate/severe-to-profound SNHL from 250 to 8000 Hz,
(dB HL)
bilaterally
SRT (dB HL)
—
—
—
—
—
50
—
S1
S2
S3
S4
S5
S6
Avg.
80
77
80
76
88
70
79
88
88
87
89
96
88
89
100
96
100
96
100
100
99
20–35 20–35 20–30 15–20 15–25 25–35
Consistent with normal-to-mild hearing loss at 500–4000
Hz
20
20
20
20
15
20
19
Note. Avg. = average; CNC = consonant–nucleus–consonant; HINT = Hearing in Noise Test; S = subject; SNHL = sensorineural hearing
loss; SRT = speech reception threshold.
4
American Journal of Audiology • 1–12
Downloaded From: http://aja.pubs.asha.org/ by a Fudan University User on 10/28/2017
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
and presented through Tucker-Davis Technologies equipment to the headphones (e.g., Koehnke & Besing, 1997,
Zheng et al., 2011, 2016).
For the AN and RT60 0.2-s conditions, the impulse
responses were measured at the eardrum with ear canal
resonance removed using KEMAR (G.R.A.S. Sound &
Vibration) in actual rooms (Besing & Koehnke, 1995; Zheng
et al., 2011). The room was rectangular with a size of 5 m
(wide) × 6 m (long) × 3 m (high). For the RT60 0.6- and
0.9-s conditions, MATLAB was used to calculate HRTFs
for each ear and each sound source location (ShinnCunningham, Desloge, & Kopco, 2001). The microphone
placement, room size, and speaker montage arrangement
were the same for all listening environments (AN and
reverberant). To achieve different RTs, different absorption coefficients were used in the MATLAB code.
To validate the comparability of these two processing
procedures, we tested two groups of listeners (10 in each
group) with NH (in AN and RT60 0.2-s environments),
with stimuli processed using impulses from KEMAR in a
real room and using HRTFs generated using MATLAB. Ten
of the subjects with NH completed all the test conditions,
and the other 10 subjects were recruited only to complete
two conditions (AN and RT0.2) for verification purposes.
Our results indicated no significant difference in localization performance between the two groups (F(1, 18) = 0.94,
p = .34), which is consistent with the results reported by
Rychtarikova et al. (2007).
Procedure
Before the experiment, an acoustic immittance screening was administered followed by a pure-tone air conduction test in a sound-treated booth for each of the subjects
with NH to make sure they had normal middle ear function
and NH sensitivity.
For listeners with BCIs, a loudness balance test was
administered to ensure equal loudness between ears. Then,
sound-field hearing tests including warble-tone thresholds
(CI-left/CI-right) and speech reception threshold (bilateral)
were administered in the same sound-treated booth to establish the appropriate signal presentation level to be used
when subjects wore both implants during testing. The processors of each BCI user were set as they were routinely
used, except that the automatic gain control (AGC) was
turned off before the experiment to avoid possible activation
of the AGC in the CI device by the experimental stimuli. This
was done because van Hoesel, Ramsden, and O’Driscoll
(2002) found that localization performance decreased
when stimuli were presented at 70 dB SPL compared with
60 dB SPL due to the activation of AGC.
All subjects were familiarized with the virtual localization test under three listening conditions: quiet/AN,
quiet /RT0.2, and SNR0/AN. The subjects were seated in a
quiet sound-treated booth, wearing circumaural earphones
(Sennheiser HD 265). The stimuli were randomly presented
from each of the nine simulated sound source locations, and
subjects received correct answer feedback. This took about
10 min. After familiarization, the range of test conditions
to be used was explored with each subject. Each subject
with BCI tried two runs in each of the four conditions
(SNR-8/RT0.9, SNR-4/RT0.9, SNR-8/RT0.6, and SNR-4/
RT0.6). These were expected to be the most difficult conditions. If the subjects tried and reported that any or all of
the conditions were too difficult, then these conditions were
not completed and the results were reported as at chance
(82° root-mean-square localization error [RMSLE]1). If the
subject did not report difficulty during testing, then the
results obtained were reported even if they exceeded chance.
Subjects with NH completed all listening conditions in a
randomized order including the most difficult conditions,
because preliminary testing with a few subjects with NH
indicated that they could localize in these conditions with
performance much better than chance.
During the experiment, the virtual localization tests
were administered to the subjects at different SNRs and
RTs in random order. A single-interval, nine-alternative,
forced-choice identification procedure with feedback was
used to measure localization ability. A computer monitor
in front of the subjects displayed a picture showing the
nine possible locations of the sound source in relation to
the subject. The subjects chose the appropriate image on
the monitor by a mouse click to indicate the perceived location of the virtual sound source.
For each subject with NH, there were 16 conditions
(4 RTs × 3 SNRs, plus the quiet condition at each RT60).
For each subject with BCI, there were 24 conditions (4 RTs ×
5 SNRs, plus the quiet condition at each RT60). In each
condition, there were 27 trials, three for each of the nine
source locations. Each condition was repeated twice, unless
the RMSLEs from these two responses differed by more than
11.25°; in these instances, the subject completed a third run.
Therefore, (unless additional runs were needed) the maximum number of stimuli presented to each subject was
16 conditions × 27 trials × 3 repetitions = 1,296 for the subjects with NH and 24 conditions × 27 trials × 3 repetitions =
1,944 for the BCI users. Each condition took 2–3 min to
complete. In total, it took about 4 hr for each subject to
complete the entire experiment, including the hearing test,
implant adjustment, training, breaks, and virtual localization tests.
Results
RMSLE in degrees was calculated for each subject
under different noise and/or reverberation conditions. The
larger the RMSLE, the poorer localization performance.
Multivariate analysis of variance and post hoc Tukey
honestly significant difference (HSD) tests were used for
1
RMSLE
was previously defined
in Zheng et al. (2016). It is given by
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
r
2ffi
∑ni¼1 xi:source − xi:response
, where xi.source and xi.response are the actual
n
sound source and the subject’s corresponding response, respectively;
their values are from the set {1, 2, 3, …, 9}; total trial, n = 27. RMS
error in degrees is the above value multiplied by 22.5.
Zheng et al.: Noise and Reverberation on BCI Sound Localization
Downloaded From: http://aja.pubs.asha.org/ by a Fudan University User on 10/28/2017
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
5
statistical analysis to determine the effects of noise and
reverberation on localization performance for each group
of subjects and to compare the performance difference
between groups. Note that the data for the quiet condition
for listeners with NH and with BCIs (in all figures in this
article) were previously published in Zheng et al. (2016)
and included in this article for comparison purposes.
Listeners With NH
The bar graph in Figure 1 shows the average RMSLE
in degrees for subjects with NH under different test conditions in both AN and reverberant environments. The acrosslistening-condition analysis for both noise and reverberation
effects showed a significant difference (noise: F(3, 36) = 13.7,
reverberation: F(3, 34) = 60.1, p < .0001), suggesting that
at least one listening condition was significantly different
from the others. Also, there was a significant interaction
between noise and reverberation effects (Noise × Reverberation: F(9, 82.898) = 4.13, p < .0001). Post hoc Tukey
HSD (α = .05) indicated that the RMSLE was significantly
larger in the −4- and −8-dB SNR conditions than in the
quiet condition and significantly larger in the RT60 0.6- and
0.9-s conditions than in the AN condition.
Results of post hoc Tukey HSD for each noise and
reverberation condition for subjects with NH are summarized
in Table 4. They indicate significantly poorer localization
performance in some noise-plus-reverberation conditions,
such as the SNR-4/RT0.6, SNR-8/RT0.6, SNR0/RT0.9,
SNR-4/RT0.9, and SNR-8/RT0.9 conditions, than in either
noise or reverberation alone for this group of subjects.
Figure 1. Sound source localization in four reverberant environments
for listeners with normal hearing. The root-mean-square localization
error (RMSLE) in degrees is indicated on the y axis as a function of
the noise level. Chance performance is indicated by the dashed line
at 82°. Standard errors are indicated for each condition. AN = anechoic;
RT = reverberation time; SNR = signal-to-noise ratio.
6
Table 4. Results of post hoc Tukey honestly significant difference
(α = .05) for each noise and reverberant condition for subjects with
normal hearing.
Listening
condition
Quiet
SNR 0 dB
SNR −4 dB
SNR −8 dB
Reverberation time
AN
▦
0.2 s
▦
▦
0.6 s
0.9 s
✰▦
✰▦
✰
✰▦
✰▦
✰▦
Note. AN = anechoic; SNR = signal-to-noise ratio; “▦” = statistically
significant difference when RT was fixed and noise condition
compared with quiet condition; “✰” = statistically significant difference
when noise condition was fixed and reverberant condition compared
with anechoic condition.
Listeners With BCIs
The bar graph in Figure 2 shows the average RMSLE
in degrees for subjects with BCI under different test conditions in both AN and reverberant environments. The acrosslistening-condition analysis for both noise and reverberation
effects showed a significant difference (noise: F(5, 30) = 70.0,
reverberation: F(3, 28) = 138, p < .0001), suggesting that
performance in at least one listening condition was significantly different from the others. Also, there was a significant interaction between noise and reverberation effects
(Noise × Reverberation: F(15, 77.697) = 4.23, p < .0001).
Post hoc Tukey HSD (α = .05) indicates a significant difference in performance between quiet and +4-, 0-, −4-, and
Figure 2. Sound source localization data in four reverberant
environments for listeners with bilateral cochlear implants. The
root-mean-square localization error (RMSLE) in degrees is
indicated on the y axis as a function of the noise level. Chance
performance is indicated by the dashed line at 82°. Standard errors
are indicated for each condition. AN = anechoic; RT = reverberation
time.
American Journal of Audiology • 1–12
Downloaded From: http://aja.pubs.asha.org/ by a Fudan University User on 10/28/2017
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
−8-dB SNR and between AN and RT60 0.2-, 0.6-, and 0.9-s
conditions for this group of subjects.
Results of post hoc Tukey HSD for each noise and
reverberant condition for the subjects with BCI are summarized in Table 5. Subjects with BCI had significantly
poorer localization performance in some noise and reverberation combined conditions, such as the SNR0/RT0.2,
SNR-4/RT0.2, SNR-8/RT0.2, SNR-4/RT0.6, SNR-8/RT0.6,
SNR + 4/RT0.9, SNR0/RT0.9, SNR-4/RT0.9, and SNR-8/
RT0.9 conditions, than in either noise or reverberation
alone.
Listeners With NH Versus Listeners With BCIs
The data were also analyzed to compare localization
performance between the two groups. Figure 3a shows
the average RMSLE in degrees of both groups of subjects
in the AN environment. Group comparison indicates that
listeners with BCIs had significantly poorer localization
accuracy than listeners with NH (group: F(1, 14) = 65.3,
p < .0001) in all conditions. The across-listening-condition
analysis showed a significant difference (noise: F(3, 12) =
15.3, p = .0002), suggesting that at least one listening condition was significantly different from the others. Also, there
was a significant interaction between listening condition and
group (Noise × Group: F(3, 12) = 6.58, p = .007). Post hoc
Tukey HSD (α = .05) indicates a significant performance
difference between the quiet and −8-dB SNR conditions
for both subjects with NH and BCI.
Figure 3b is similar to Figure 3a but for the RT60
0.2-s reverberant environment. The data also show a significant difference between groups (group: F(1, 14) = 203,
p < .0001) and across listening conditions (noise: F(3, 12) =
118, p < .0001) and a significant interaction between listening condition and group (Noise × Group: F(3, 12) = 54.7,
p < .0001). Tukey HSD test for the data of the group with
NH indicates a significant difference between quiet and
−4- and −8-dB SNR conditions. For the group with BCI,
the significant difference occurred between quiet and 0-, −4-,
and −8-dB SNR conditions.
Figure 3c shows the data for the RT60 0.6-s reverberant environment. There is a significant difference between
groups (group: F(1, 14) = 151, p < .0001) and across listening conditions (noise: F(3, 12) = 25.6, p < .0001). However,
there is no significant interaction between listening condition and group (Noise × Group: F(3, 12) = 0.30, p = .83).
Tukey HSD results indicate a significant difference between
quiet and −4- and −8-dB SNR conditions for both groups
of subjects (with NH and with BCI).
Figure 3d shows the data for the RT60 0.9-s reverberant environment. The results also reveal a significant difference between groups (group: F(1, 14) = 214, p < .0001) and
across listening conditions (noise: F(3, 12) = 69.0, p < .0001).
However, there is no significant interaction between listening condition and group (Noise × Group: F(3, 12) = 2.07,
p = .16). Tukey HSD results indicate a significant difference
between quiet and 0-, −4-, and −8-dB SNR conditions for
subjects with NH and between quiet and +4-, 0-, −4-, and
−8-dB SNR conditions for subjects with BCI.
Figure 3 reveals the localization performance difference between the two groups of subjects. In addition, it
shows the performance pattern as noise increases in each
reverberant environment. That is, localization error increases
as SNR decreases. There were different localization performance patterns in AN and RT0.2 environments for the
two groups of subjects; as the noise level increases, there
was a greater effect on listeners with BCIs than those with
NH, but in RT0.6 and RT0.9 environments, performance
was similar.
Among the six subjects with BCI, two used devices
from the Advanced Bionics Corporation (ABC); and four,
from Cochlear Corporation (CC). Overall, there was no
clear or consistent localization performance difference across
the subjects in different listening environments using different devices. However, no statistical analysis was possible
due to limited sample size.
Discussion and Conclusion
Listeners With NH
Table 5. Results of post hoc Tukey honestly significant difference
(α = .05) for each noise and reverberant condition for subjects with
bilateral cochlear implants.
Listening
condition
Quiet
SNR +8 dB
SNR +4 dB
SNR 0 dB
SNR −4 dB
SNR −8 dB
Reverberation time
AN
0.2 s
✰
▦
✰▦
✰▦
✰▦
0.6 s
0.9 s
✰
✰
✰
✰
✰▦
✰▦
✰
✰
✰▦
✰▦
✰▦
✰▦
Note. AN = anechoic; SNR = signal-to-noise ratio; “▦” = statistically
significant difference when RT was fixed and noise condition compared
with quiet condition; “✰” = statistically significant difference when
noise condition was fixed and reverberant condition compared with
anechoic condition.
Results for listeners with NH revealed a significant
effect of noise and reverberation on localization performance.
Figure 1 clearly indicates that localization accuracy decreases
as SNR decreases and as reverberation increases. The significant change in RMSLE occurred at a −4-dB SNR and
0.6-s RT60. These results are consistent with previous reports
(e.g., Giguere & Abel, 1993; Lorenzi et al., 1999a; Zheng
et al., 2011).
In everyday listening situations, both noise and reverberation are present. Therefore, it is important to consider
the combined effects of noise and reverberation on localization ability. Table 4 lists the five listening conditions in
which noise and reverberation combined had a significant
effect on localization ability: SNR-4/RT0.6, SNR-8/RT0.6,
SNR0/RT0.9, SNR-4/RT0.9, and SNR-8/RT0.9. That is,
noise plus reverberation resulted in significantly poorer localization performance at lower SNRs and shorter RT60s than
Zheng et al.: Noise and Reverberation on BCI Sound Localization
Downloaded From: http://aja.pubs.asha.org/ by a Fudan University User on 10/28/2017
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
7
Figure 3. Sound source localization data in anechoic and reverberant environments for listeners with NH and BCIs. The root-mean-square
localization error (RMSLE) in degrees is indicated on the y axis versus the listening conditions on the x axis. Chance performance is indicated
by the dashed line at 82°. Standard errors are indicated for each condition for subjects with NH. AN = anechoic; BCI = bilateral cochlear
implant; NH = normal hearing; RT = reverberation time.
either noise or reverberation alone. The post hoc Tukey
HSD results for listeners with NH indicate a sequential decrease in localization performance under different listening
conditions, with the poorest performance in the SNR-8/
RT0.9 condition, followed by the SNR-4/RT0.6, SNR0/
RT0.9, then SNR0/RT0.6, SNR-4/RT0.2, SNR-8/RT0.2,
SNR-8/AN, SNR-4/AN, quiet/RT0.6, and quiet/RT0.9
conditions, and the best performance in the quiet/AN, quiet/
RT0.2, SNR0/AN, and SNR0/RT0.2 conditions. This confirms that there is a greater effect of noise plus reverberation
on localization performance for individuals with NH than
noise or reverberation alone.
Listeners With BCIs
Results for listeners with BCIs also reveal a significant
effect of noise and reverberation on localization. Figure 2
indicates that localization accuracy decreases as SNR
decreases and as reverberation increases. The significant
change in RMSLE occurred at +4-dB SNR and 0.2-s RT60,
respectively, for the group with BCI. That is, RMSLE was
significantly larger at +4-dB SNR than +8-dB SNR and was
significantly larger at 0.2-s RT60 than in the AN environment.
These results are consistent with previous reports (Kerber
& Seeber, 2013; Zheng et al., 2011). The localization results
8
for subjects with BCI in quiet were consistent with those
obtained by Neuman et al. (2007), but the RMSLEs were
larger than those measured by Grantham et al. (2007). This
may be due to the different measurement methods used in
the studies. In Grantham et al., stimuli were presented from
17 loudspeakers in the frontal horizontal plane (±80°), 10°
apart. However, Neuman et al. used nine speakers in the
sound field in the frontal horizontal plane (±90°), 22.5° apart.
This study used a source arrangement similar to Neuman
et al., except that we used virtual source locations and we did
not use a rove to remove monaural cues. The smaller distance
between adjacent speakers in Grantham et al.’s study may
have resulted in smaller RMSLEs. In addition, subjects in
Grantham et al.’s study used different devices (Med-EL) and
were slightly younger (mean age = 47.6 years) than those in
Neuman et al.’s study (Nucleus; mean age = 51.6 years) and
this study (Nucleus and ABC; mean age = 58.5 years), which
may explain some of the performance difference. Also,
individual differences may account for the performance variability across studies, due to different etiologies of hearing
loss, a wide range of residual capability of the auditory system, and auditory experience.
For the 0-dB SNR condition (the only condition
common to both Van Hoesel et al. [2008] and this study),
the RMSLEs derived in this study were consistent with those
American Journal of Audiology • 1–12
Downloaded From: http://aja.pubs.asha.org/ by a Fudan University User on 10/28/2017
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
measured by Van Hoesel et al. (2008). Both Van Hoesel
et al. and Neuman et al. tested subjects in the sound field;
the fact that our results are comparable with theirs further
confirms the feasibility of using this virtual localization test
with BCI users. The data obtained in this study clearly indicate that virtual localization tests are an effective clinical
tool for evaluating BCI users’ binaural performance.
The results of this study confirm and extend our previous study regarding the effects of duration of experience
with BCIs on localization performance. Two subjects with
BCI were tested by Zheng et al. (2011) using the virtual
localization experiment in quiet and noise (SNR = 4, 0,
−4, and −8 dB) in AN and 0.2-s RT environments. In this
study, the same two subjects were tested again (18 months
later) using the same virtual localization test in the same
AN and 0.2-s RT conditions; they were also tested with
longer RTs (RT60 = 0.6 and 0.9 s) and in an additional noise
condition (+8-dB SNR). Figure 4 shows the average RMSLE
in degrees for subjects with BCI under different test conditions in AN and 0.2-s RT60 reverberant environments in
the previous and current studies. Both of the BCI users in
Zheng et al. (2011) had less than 1 year of BCI experience;
in this experiment, both subjects with BCI had 2 or more
years of BCI experience. Results indicate that there is a
clear improvement in localization performance with increasing BCI experience in most listening conditions. Interestingly,
in very noisy environments (SNR-8/AN and SNR-8/RT0.2),
there was no clear difference between the localization performance in Zheng et al. (2011) and that in the current study,
as localization performance was equally poor in both
Figure 4. Sound source localization data from the 2011 study and
the current study in two reverberant environments (AN and 0.2-s
RT60) for listeners with BCIs. The root-mean-square localization
error (RMSLE) in degrees is indicated on the y axis as a function
of the noise level. Chance performance is indicated by the dashed
line at 82°. Standard errors are indicated for each condition. AN =
anechoic; BCI = bilateral cochlear implant; RT = reverberation time.
instances. Also, in the SNR-4/RT0.2 condition, the RMSLE
in the current study was slightly larger than that in the
previous study (Zheng et al., 2011). Research studies have
indicated that localization performance in quiet stabilizes
after a year of BCI experience (e.g., Tyler, Noble, Dunn, &
Witt, 2006). However, typical listening environments contain both noise and reverberation, which degrade the interaural cues used for sound localization. It is not clear how
experience will affect BCI users’ localization performance
in everyday listening situations. Only two of our subjects
participated in our initial study (Zheng et al., 2011) and in
this study; therefore, more data are needed from additional
subjects to conclusively determine the effect of the duration
of BCI experience on localization performance in adverse listening environments containing both noise and reverberation.
It is possible that, in difficult noisy and reverberant environments, there will be no improvement in localization ability
even with more experience, due to limitations in the devices
and/or the signal processing strategies.
In this study, we focused on examining the combined
effect of noise and reverberation. Table 5 lists the nine
listening conditions in which there was a significant combined effect of noise and reverberation on localization
ability: SNR4/RT0.9, SNR0/RT0.2, SNR0/RT0.9, SNR-4/
RT0.2, SNR-4/RT0.6, SNR-4/RT0.9, SNR-8/RT0.2, SNR-8/
RT0.6, and SNR-8/RT0.9. That is, noise plus reverberation
resulted in significantly poorer localization performance at
higher SNRs and shorter RT60s than either noise or reverberation alone. Binaural cues that are important for sound
localization, including ILDs and ITDs, are both affected
by noise and reverberation. Competing noise masks the
signal and reduces ILDs. Reflected sounds due to reverberation may degrade the original sound, reducing ILDs and
disrupting ITDs depending on the source location and the
actual RT. A study of 22 BCI users with postlingual deafness
by Grantham et al. (2007) found that ILD is the dominant
cue for sound localization but ITDs may contribute to localization of speech based on available timing cues in the speech
envelope. This suggests that noise plus reverberation will
degrade binaural cues more and result in poorer localization
than either noise or reverberation alone. A typical classroom/conference room has an SNR of about +5 dB and
an RT60 of around 0.2–0.4 s. Results of this study indicate
that BCI users are likely to experience localization difficulty in such an environment. Post hoc Tukey HSD results
clearly indicate that localization error increases as the listening condition becomes more adverse; this confirms that
the combined effect of noise and reverberation on the localization performance of BCI users is greater than the effect
of either factor alone.
Six BCI users were tested in this study; two of them
used ABC devices, and four used CC devices. There were
no clear or consistent differences seen in localization performance among these six BCI users. This is not surprising
given the small standard error obtained for the group with
BCI overall. One explanation might be the homogeneity of
this group of BCI users; all of them had good speech understanding in quiet. Table 3 indicates that all six subjects with
Zheng et al.: Noise and Reverberation on BCI Sound Localization
Downloaded From: http://aja.pubs.asha.org/ by a Fudan University User on 10/28/2017
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
9
BCI had good-to-excellent scores on consonant–nucleus–
consonant words, consonant–nucleus–consonant phonemes,
and Hearing in Noise Test in quiet with their CIs. On the
basis of patient reports, they all had very good speech understanding in daily listening situations. Both speech understanding, especially in noise, and sound localization rely on
binaural cues. In addition, speech perception depends on
the resolution of temporal cues, which are important for
sound localization, especially in a reverberant environment.
Therefore, speech and localization performance may exhibit
similar effects. That is, speech intelligibility and localization
performance in individuals with BCIs may be affected similarly by noise and reverberation. More data are needed to
address this topic.
Listeners With NH Versus BCIs
Compared with listeners with NH, BCI users have
significantly poorer localization performance in all virtual
environments. BCI users’ localization performance started
to decrease at a higher SNR (NH: −4 dB, BCI: +4 dB)
and a shorter RT60 (NH: 0.6 s, BCI: 0.2 s) than listeners
with NH. In addition, the combined effect of noise plus
reverberation occurred at a higher SNR and a shorter RT60
(SNR + 4/RT0.9 or SNR0/RT0.2) for BCI users than for
listeners with NH (SNR0/RT0.9 or SNR-4/RT0.6). These
results indicate that, although bilateral implantation does
improve spatial hearing, it does not completely restore localization ability. This is consistent with previous studies
described in Preece (2010). There are two primary factors
responsible for the poorer-than-normal localization performance for listeners with BCIs. One is from the CI device
itself. It has a finite number of electrodes, and the speech
processor also has limited capabilities. Current CIs have
from 16 (ABC) to 22 (CC) and 24 (Med-EL) electrodes
(Wolfe & Schafer, 2015). Maximum channel stimulation
varies with different speech processing strategies even for
the same device and, depending on the electrode placement
and number of electrodes, often does not fully represent
the frequency range of the cochlea. Also, surgical insertion
of the electrodes in the cochlea in each ear might be slightly
different due to anatomical difference between ears, resulting in reduced interaural correlation of signals. CI signal
processing strategies attempt to preserve the temporal fine
structure of the signals by increasing the number of channels. However, increasing the stimulation rate may reduce
ITD sensitivity due to interference from adjacent electrodes
(Jones, Litovsky, & Van Hoesel, 2007). Grantham et al.
(2007) and others have found that ILDs are the dominant
cue for sound localization for BCI users. In addition, the
difference between the input dynamic range in the speech
processor and electrical stimulation further reduces the
interaural correlation causing decreasing binaural benefit
(Zeng & Galvin, 1999). The other factor responsible for
poorer localization may be the characteristic of the implant
users. Many BCI users may not use the binaural cues to the
extent listeners with NH use them due to limited residual
auditory function (Shannon, Fu, Galvin, & Friesen, 2004).
10
All these factors result in reduced binaural benefit and
poorer localization performance. The findings of this study
indicate the need to improve noise and reverberation reduction algorithms for CIs to help CI users detect and localize
sounds better in everyday listening situations. The subjects
recruited for this study were all highly proficient BCI users.
The information obtained in this study should be useful in
the design of future CIs as these data indicate the amount
of noise and reverberation that can be tolerated by CI users
before spatial hearing ability is degraded.
Interestingly, there was a significant interaction between listening condition and group for AN and 0.2-s RT60
reverberant environments but no such interaction for RT60
of 0.6- and 0.9-s reverberant environments (see Figure 3).
This means that, in 0.6- and 0.9-s RT60 environments, both
groups of listeners show a similar trend in performance
across noise conditions. Figure 3 shows localization performance for both groups of subjects across listening environments. In the AN environment, although both groups
of subjects had significantly larger localization errors in
the −8-dB SNR condition than in the quiet condition, the
−8-dB SNR condition had a significantly greater effect
on the localization performance of subjects with BCI than
on the localization performance of subjects with NH. In the
0.2-s RT60 environment, BCI users’ performance started to
decrease at a higher SNR (0 dB) than listeners with NH
(−4 dB). The slopes of the localization error functions across
noise condition appear steeper for subjects with BCI than
subjects with NH in both AN and 0.2-s RT environments.
However, in the 0.6- and 0.9-s RT60 environments, the
slopes of the functions for both groups of subjects were nearly
parallel to one another and showed the same pattern of
change as a function of noise condition. This finding may
reveal the fact that the normal auditory system has the
ability to tolerate the effects of noise and reverberation to
some extent. Once RT reaches a certain point, it has a comparable effect on all listeners. These data further confirm
the need for better noise and reverberation reduction algorithms in CI devices to improve CI users’ spatial hearing in
everyday listening situations.
Summary and Conclusions
The results of this study reveal that noise and reverberation have a significant effect on the localization performance of listeners with NH and BCIs but the effect on
BCI users’ performance is greater. For both groups of listeners, localization performance decreases as SNR decreases
and as RT increases. Compared with listeners with NH,
BCI users’ localization performance starts to decrease at a
higher SNR and at a shorter RT60 for noise and reverberation alone and combined.
This study investigated the combined effects of noise
and reverberation in listeners with NH and BCIs. The data
confirm that the virtual localization test used in this study
is effective for measuring binaural performance in listeners
with BCIs and sensitive to performance differences between
the two groups of subjects (with NH and with BCIs; Zheng
American Journal of Audiology • 1–12
Downloaded From: http://aja.pubs.asha.org/ by a Fudan University User on 10/28/2017
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
et al., 2011). Validating the use of virtual tests will likely be
of great benefit in the audiology clinic. One benefit is that
virtual simulation makes it possible to approximate real-life
listening situations with various levels of noise and reverberation without the necessity of using different rooms and
additional equipment. Also, the virtual test makes it possible
to avoid problems faced in sound-field testing such as head
movement, calibration, and difficulty varying RT (ASHA,
1991). Another important benefit is to the patients: The virtual test may be used as a measure to evaluate changes in
localization ability over time, with or without specific intervention. Audiologists may use stimuli such as those in this
study with their patients to practice localizing in different
noise and reverberation conditions; it is possible that localization performance will improve over time. Also, a questionnaire may be developed to accompany these virtual stimuli
to evaluate patients’ perceived benefit from practicing with
the virtual stimuli in typical listening situations. Future studies are planned to address this use of the virtual stimuli in
this study.
Acknowledgments
This research was funded by a grant from the American
Speech-Language-Hearing Foundation and partially supported
by new faculty funds from Central Michigan University. The
data for the quiet condition from Zheng et al. (2016) were included in this article for comparison purposes. Many thanks to
all the subjects for their participation in this research project.
References
American Speech-Language-Hearing Association. (1991). Sound
field measurement tutorial. Asha, 33(Suppl. 3), 25–37.
Besing, J. M., & Koehnke, J. (1995). A test of virtual auditory
localization. Ear and Hearing, 16, 220–229.
Gelfand, S. A. (2009). Hearing: An introduction to psychological
and physiological acoustics (5th ed.). New York, NY: Taylor &
Francis Group.
Giguere, C., & Abel, S. M. (1993). Sound localization: Effects
of reverberation time, speaker array, stimulus frequency, and
stimulus rise/decay. The Journal of the Acoustical Society of
America, 94, 769–776.
Good, M. D., & Gilkey, R. H. (1996). Sound localization in noise:
The effect of signal-to-noise ratio. The Journal of the Acoustical Society of America, 99, 1108–1117.
Grantham, D. W., Ashmead, D. H., Ricketts, T. A., Labadie, R. F.,
& Haynes, D. S. (2007). Horizontal-plane localization of noise
and speech signals by postlingually deafened adults fitted with
bilateral cochlear implants. Ear & Hearing, 28, 524–541.
Hartmann, W. M. (1983). Localization of sound in rooms. The
Journal of the Acoustical Society of America, 74, 1380–1391.
Jones, G., Litovsky, R., & Van Hoesel, R. (2007). ITD sensitivity in
electrical hearing: Effect of channel interactions. CIAP Meeting,
Lake Tahoe, NV.
Kerber, S., & Seeber, B. U. (2012). Sound localization in noise
by normal-hearing listeners and cochlear implant users. Ear &
Hearing, 33, 445–457.
Kerber, S., & Seeber, B. U. (2013). Localization in reverberation
with cochlear implants: Predicting performance from basic
psychophysical measures. Journal of the Association for Research
in Otolaryngology, 14, 379–392.
Koehnke, J., & Besing, J. (1997). Clinical applications of 3-D
auditory tests. Seminars in Hearing, 18, 345–354.
Kopco, N., Best, V., & Shinn-Cunningham, B. G. (2007). Sound
localization with a preceding distractor. The Journal of the
Acoustical Society of America, 121, 420–432.
Lorenzi, C., Gatehouse, S., & Lever, C. (1999a). Sound localization
in noise in normal-hearing listeners. The Journal of the Acoustical Society of America, 105, 1810–1820.
Lorenzi, C., Gatehouse, S., & Lever, C. (1999b). Sound localization
in noise in hearing-impaired listeners. The Journal of the Acoustical Society of America, 105, 3454–3463.
Mosnier, I., Sterkers, O., & Bebear, J. P. (2009). Speech performance and sound localization in a complex noise environment
in bilaterally implanted adults. Audiology & Neuro-Otology,
14, 106–114.
Neuman, A., Haravon, A., Sislian, N., & Waltzman, S. (2007).
Sound-direction identification with bilateral cochlear implants.
Ear and Hearing, 28, 73–82.
Preece, J. P. (2010). Sound localization by cochlear implant users.
Seminars in Hearing, 31, 37–46.
Rychtarikova, M., Van den Bogaert, T., Vermeir, B., & Wouters, J.
(2007). Virtual acoustics for localization of the speaker in a real
and virtual reverberation room. Paper presented at Inter-Noise
2007, Istanbul, Turkey.
Schön, F., Müller, J., Helms, J., & Nopp, P. (2005). Sound localization and sensitivity to interaural cues in bilateral users of
the Med-El Combi 40/40 + cochlear implant system. Otology
& Neurotolology, 26, 429–437.
Shannon, R. V., Fu, Q., Galvin, J., & Friesen, L. (2004). Chapter 8:
Speech perception with cochlear implants. In G. Zeng, A. N.
Popper, & R. R. Fay (Eds.), Cochlear implants: Auditory prostheses and electric hearing (pp. 334–376). New York, NY:
Springer.
Shinn-Cunningham, B. G., Desloge, J. G., & Kopco, N. (2001).
Empirical and modeled acoustic transfer functions in a simple
room: Effects of distance and direction. Paper presented at
2001 IEEE Workshop on Applications of Signal Processing
to Audio and Acoustics, New Paltz, NY.
Tyler, R. S., Noble, W., Dunn, C., & Witt, S. (2006). Some benefits and limitations of binaural cochlear implants and our
ability to measure them. International Journal of Audiology,
45(Suppl. 1), S113–S119.
Van Hoesel, R., Bohm, M., Pesch, J., Vandali, A., Battmer, R. D.,
& Lenarz, T. (2008). Binaural speech unmasking and localization in noise with bilateral cochlear implants using envelope
and fine-timing based strategies. The Journal of the Acoustical
Society of America, 123, 2249–2263.
Van Hoesel, R., Ramsden, R., & O’Driscoll, M. (2002). Sounddirection identification, interaural time delay discrimination, and speech intelligibility advantages in noise for
a bilateral cochlear implant user. Ear & Hearing, 23(2),
137–149.
Van Hoesel, R. J. M., & Tyler, R. S. (2003). Speech perception,
localization, and lateralization with bilateral cochlear implants. The Journal of the Acoustical Society of America, 113,
1617–1630.
Verschuur, C. A., Lutman, M., Ramsden, R., Greenham, P., &
O’Driscoll, M. (2005). Auditory localization abilities in bilateral cochlear implant recipients. Otology & Neurotology, 26,
965–971.
Wolfe, J., & Schafer, E. C. (2015). Programming cochlear implants
(2nd ed.). San Diego, CA: Plural Publishing, Inc.
Zheng et al.: Noise and Reverberation on BCI Sound Localization
Downloaded From: http://aja.pubs.asha.org/ by a Fudan University User on 10/28/2017
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
11
Zeng, F. G., & Galvin, J. J. (1999). Amplitude compression and
phoneme recognition in cochlear implant listeners. Ear &
Hearing, 20, 60–74.
Zheng, Y., Koehnke, J., & Besing, J. (2016). Effects of reverberation on sound localization for bilateral cochlear implant users.
12
Journal of Phonetics and Audiology, 2, 108. https://doi.org/
10.4172/jpay.1000108
Zheng, Y., Koehnke, J., Besing, J., & Spitzer, J. (2011). Effects of
noise and reverberation on virtual sound localization for listeners
with bilateral cochlear implants. Ear & Hearing, 32, 569–572.
American Journal of Audiology • 1–12
Downloaded From: http://aja.pubs.asha.org/ by a Fudan University User on 10/28/2017
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
Документ
Категория
Без категории
Просмотров
3
Размер файла
475 Кб
Теги
2017, aja, 0101
1/--страниц
Пожаловаться на содержимое документа