2019 Conference on Implantable Auditory Prostheses

14-19 July 2019

Granlibakken, Lake Tahoe

Page 249

W19: DECODING THE AUDITORY NERVE.

IMPLICATIONS FOR COCHLEAR IMPLANT TECHNOLOGY

Jacques Andre Grange, John Francis Culling

Cardiff University, United Kingdom, Cardiff, GBR

A vocoder based on the MAP-simulated (Model of the Auditory Periphery, e.g. Meddis, 2006)

spiking of 30,000 auditory nerve fibers led to near-normal speech intelligibility of vocoded

speech. The vocoder thereby reveals a very simple and computationally efficient way to predict

how AN firing patterns and their encoding of speech waveform temporal fine structure (TFS)

may be interpreted by the brain.

TFS transmission through cochlear-

implants (CIs) is essential to improving CI users’ pitch

perception, music perception/appreciation and, of course, speech intelligibility. A breakthrough

in CI TFS transmission is yet to be found. Part of the reason for this is that focus has mostly

been placed on improving spectral resolution at the AN excitation level, which is inherently

limited by spread of excitation. The proposed encoder predicts that, in response to a speech

stimulus, the AN encodes signal periodicity (fundamental frequency and formants) and temporal

modulations using spike timing. Physiological evidence (e.g. Shamma et al., 1985) suggests

that such information is transmitted over as few as five, time-varying spectral bands. If so, the

spectral smearing effect of CI spread of excitation would become a lot easier to mitigate. We

demonstrate, here, how the vocoder can be used to evaluate the potential benefit of different

fine-structure encoding strategies. Overall, the above approach may facilitate the development

of real-time, population-level predictions of the effectiveness of AN-encoding, as well as help

determine the likelihood of success in TFS transmission in CIs.