Using Acoustic Models for Sound Localization with

1
Using Acoustic Models for Sound Localization with
Bilateral Cochlear Implants Nadeem R. Kolia,
Joshua S. Stohl, Philip R. Brown, Debara L.
Tucci, and Leslie M. Collins Department of
Electrical and Computer Engineering, Duke
University, Durham, NC
Introduction Bilateral cochlear implants are
increasingly common among cochlear implant users,
as studies have shown that bilateral cochlear
implants have the potential to improve
performance in sound localization and hearing in
noise tasks (van Hoesel and Tyler, 2003 Litovsky
et al, 2004). While it has been established that
using acoustic models and normal hearing
listeners is a valid method for predicting trends
in unilateral implants, it has not yet been
demonstrated that the same is true for bilateral
cochlear implants. The goal of this study was to
establish the validity of binaural acoustic
models for predicting performance of bilateral
cochlear implant users on a sound localization
task. A bilateral cochlear implant sound
localization experiment with adults (Litovsky et
al, 2004) was replicated with normal hearing
listeners by using a virtual acoustic source
synthesized using digital filters constructed
from head-related transfer functions (HRTFs) to
model the free-field stimuli used by Litovsky et
al, (2004). A secondary objective of this study
was to explore how multi-rate algorithms (Nie et
al., 2005 Throckmorton et al., 2006) might
improve sound localization performance with
bilateral cochlear implants. Results suggest that
binaural acoustic simulations of bilateral
cochlear implants may be useful in predicting
trends in cochlear implant sound localization
ability. It is unclear whether or not multi-rate
sound processing strategies provide any
additional benefit in a sound localization task.
Methods Task 8 subjects were asked to identify
the angle of the sound source Setup 8 virtual
speakers in a semicircular array -70 to 70 from
the azimuth. Virtual location was modeled using
HRTFs (See next section) generated from
measurements recorded with the Knowles Electronic
Manikin for Acoustic Research (KEMAR). Trials
For each combination of condition (Left, Right,
Binaural) and strategy (Unprocessed, ACE, FAME,
MCFA), 20 stimuli were presented from each
speaker Stimuli 4 bursts of 170 ms of pink
noise, with 10 ms rise/fall times, and 50 ms
between bursts were presented at 65 dB and
randomly roved by 6 dB SPL.
Sound Processing Strategies The majority of
clinical sound processing strategies are based on
the Continuous Interleaved Sampling (CIS)
strategy. In CIS, input is passed through a bank
of band pass filters. The envelope of each
subband is used to amplitude modulate a biphasic
pulse train emitted from the associated
electrode. The N-of-M strategy builds upon CIS by
only exciting the N electrodes whose subbands
have the most energy. Both N-of-M and CIS are
fixed rate strategies the rate of the biphasic
pulse train is fixed, and only the amplitude is
allowed to vary. Frequency Amplitude Modulation
Encoding (FAME) and Multi-Carrier Frequency
Modulation Algorithm (MCFA) are multi-rate
strategies. In FAME, the rate of the biphasic
pulse train is allowed to vary continuously
within each subband according to instantaneous
frequency. MCFA differs from FAME by quantizing
the stimulation rates assigned to each subband to
predefined values.
HRTF The Head Related Transfer Function (HRTF) is
a digital filter which encompasses all three
sound localization cues. It can be used to
process sound so that it appears to come from a
virtual source location when played back through
stereo headphones.
Results
Sound Localization Subjects were asked to perform
a sound localization task. Three cues are used by
the brain to localize sound Interaural Time
Difference (ITD) is the difference in the time
the sound takes to reach each ear. Interaural
Level Difference (ILD) is the difference in the
intensity of the sound reaching each ear.
Filtering Effects The properties of the head and
ears frequency filter sound waves depending on
the angle of incidence.
Pitch Perception Two factors contribute to how
pitch is perceived. According to Place Theory,
pitch is perceived as a function of the location
of excitement along the basilar membrane. In Rate
Theory, the firing rate of the neurons is
responsible for the perceived pitch. Clinical
sound processing strategies used today rely only
upon the place theory of pitch perception. Newer
multi-rate strategies hope to improve performance
by also taking advantage of Rate Theory.
Left Mean and standard deviation
root-mean-square (RMS) error for different
strategies and conditions, including data from
Litovsky et al., 2004. Center, Right Confusion
matrices for two subjects. Middle subject shows
good performance in all conditions and
strategies. Right subject does poorly in the
monaural condition but does well in the binarual
condition.
Discussion In each of the 3 conditions, no
significant difference was observed between
acoustic and electric stimulation results. Using
a one-way analysis of variance, significant
improvement was observed for binaural hearing
over both monaural hearing conditions for every
strategy (plt0.02). Although not significant in
the Litovsky et al (2004) study, a similar
improvement was observed. No significant
difference was seen across strategy within any
condition. Based on these data, acoustic models
can be used for predicting performance of
bilateral cochlear implant users in sound
localization tasks. Sound processing strategy
does not appear to have a significant effect on
sound localization performance, and multi-rate
strategies do not appear to offer an advantage
over single-rate strategies. These sound
processing strategies mainly operate to extract
spectral information, while leaving temporal
information and intensity information relatively
intact, thus ITD and ILD cues may not be
significantly affected by sound processing
strategy.
Litovsky, R. Y. Parkinson, A. Arcaroli, J.
Peters, R. Lake, J. Johnstone, P. Yu, G.
(2004), 'Bilateral Cochlear Implants in Adults
and Children', /Arch Otolaryngol Head Neck Surg/
130, 648-655. Nie, K. Stickney, G. Zeng, F.
(2005), 'Encoding frequency modulation to improve
cochlear implant performance in noise', IEEE
Transactions on Biomedical Engineering 52(1),
64-73. Throckmorton, C.S. Kucukoglu, M.S.
Remus, J.J. Collins, L. (2006), 'Acoustic model
investigation of a multiple carrier frequency
algorithm for encoding fine frequency structure
Implications for cochlear implants', Hearing
Research 218(1-2), 30-42. 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(3),
1617-1630.