Perceptual adaptation to a binaurally mismatched frequency to place map
Document Sample


Perceptual adaptation to a binaurally mismatched frequency-to-place map
Catherine Siciliano, Andrew Faulkner, Stuart Rosen
Department of Phonetics and Linguistics and UCL Centre for Human Communication
UCL, London, UK
Background Can listeners adapt? Results Conclusions
The cochlea analyzes incoming sound into its frequency Simulations of cochlear implants were used to test whether IEEE sentence intelligibility The findings suggest that listeners rely on the natural
components and passes this information on to higher centres normal-hearing listeners could learn to integrate mismatched 100 frequency-to-place map when it is available, even at the
in the brain. A cochlear implant is an implantable device that frequency maps. Processing was designed to ensure that the expense of improved spectral detail. The perceptual
can restore functionality to faulty cochleae. The implant integration of information from both ears would be maximally adaptation must involve a rapid attuning to just the matched
performs the frequency analysis and passes this information beneficial. 80 frequency bands - subjects are not simply listening to the
on to the brain via direct electrical stimulation. better ear.
Historically, cochlear implantees have been fitted unilaterally
with a single implant. However, research has shown that
60
condition
Hypothesis 1:
hearing with two ears is better than hearing with a single ear, Subjects may need a period of training with the shifted map in
% keywords correct
especially in background noise. This is especially true for dichotic unshifted the absence of the natural frequency-place map to realize any
40
cochlear implantees. Accordingly, implant patients are now dichotic part-shifted improvements with the mismatched map. Experiments testing
sometimes fitted with a hearing aid in the opposite ear to make this hypothesis are currently underway.
use of what residual hearing may be left. Some patients have 20 diotic part-shifted
been fitted with two cochlear implants. In both cases, benefits 3 unshifted channels Hypothesis 2:
to speech perception have been demonstrated. [1,2,3] Performance never reached asymptotic level, so the training
0 3 shifted channels period may not have been long enough for subjects to adapt to
Spectral shifts 0 4 8
the processor. Listeners may outperform the 3 matched
The snail-shaped anatomy of the cochlea can prevent the training sessions (40 minutes each) channels after very long-term training with the mismatched
cochlear implant electrode array from being fully inserted. processor.
Consequently, the cochlear implant may deliver its frequency Repeated-measures ANOVA:
analysis to the 'wrong' place along the basilar membrane. No significant interaction between talker and condition Implications
Experimental processor: 6-channel dichotic sine-vocoder with
We can liken this to an upward spectral shift of frequency Significant training effect (p < 0.05) for all conditions While listeners can learn to adapt to very distorted speech,
a 6mm shift in one ear [6]
information. Shallow implant insertions have been shown to Bonferroni-adjusted paired comparisons (post-training): they are quicker and better at adapting when this speech
have adverse effects on speech perception immediately Subjects retains the natural relative frequency order (cf. [4]). This
following surgery. Over time, though, many listeners can learn No significant difference between dichotic part-shifted,
should be taken into account when optimizing bilateral devices
Six normally-hearing native speakers of British English diotic part-shifted and 3 unshifted channels (p > 0.05)
to accommodate this 'spectral shift.' [4] for speech recognition.
Previous research has examined spectral shifts that are both Training Vowel identification
unilateral and uniform. Bilateral devices pose new questions 100
10 minute familiarization block with unshifted processor
about spectral shifts as frequency maps are likely to be
asymmetric. Dichotic part-shifted processor References
Connected Discourse Tracking (CDT) 80 [1] Ching, Y. C. et al. (2001). Ear & Hear 22(5), 365-380.
8 40-minute training sessions [2] Tyler, R. et al. (2002). Ear & Hear 23(1), 80S-89S.
5 minutes per session with visual feedback [3] Dorman, M. & Dahlstrom, L. (2004) Ear & Hear 25, 191-194.
60 condition [4] Rosen, S., Faulkner, A. & Wilkinson, L. (1999). JASA 106, 3629-
Testing dichotic unshifted 3636.
Pre-test, mid-test and post-test 40
% vowels correct
[5] Images adapted from Geisler, C.D. (1998). From Sound to
dichotic part-shifted Synapse: Physiology of the Mammalian Ear.
apex
base IEEE/Harvard sentences
diotic part-shifted [6] Greenwood, D. D. (1990) JASA 87(6), 2592-2605.
The apex is impenetrable by the implant. Low Vowel identification: 'Say bVd again.' 20
frequency information important for speech is
presented to structures further towards the base - cochlear implant 1 male, 2 female talkers 3 unshifted channels
a 'basalward' shift. electrode
0 3 shifted channels Acknowledgments
An unwound cochlea fitted with an implant (spectral shift)
0 4 8 Work supported by EU project HEARCOM [FP6-004171].
condition right ear left ear
The base of the cochlea is most responsive to high-frequency training sessions (40 minutes each)
sounds. Here the inner hair cells (nerve receptors) have died, and dichotic unshifted 1, 3, 5 2, 4, 6
the cochlea is unresponsive.
Repeated-measures ANOVA:
dichotic part-shifted 1, 3, 5 à 6mm 2, 4, 6
No significant interaction between talker and condition
1, 3, 5 à 6mm 1, 3, 5 à 6mm Significant training effect (p < 0.05) for all conditions
diotic part-shifted
base apex
2, 4, 6 2, 4, 6 Bonferroni-adjusted paired comparisons (post-training):
Surviving hair cells at the
apex respond to low
3 unshifted channels 2, 4, 6 Performance with 3 matched channels better than with
frequencies with the help dichotic and diotic part-shifted conditions (p < 0.01)
of a hearing aid. 3 shifted channels 1, 3, 5 à 6mm
An unwound cochlea in the opposite ear (no spectral shift) [5]
Related docs
Get documents about "