Examining spectral and intensity resolution in web-based vocoder simulations of electric hearing

Abstract

The poor spectral resolution in current cochlear implants contributes to the difficulty implant users face when trying to listen in noisy backgrounds. Recent work examining measures of spectral resolution and speech perception in normal-hearing and cochlear-implant listeners found that the amount of simulated spectral resolution producing similar speech-in-noise performance between groups resulted in poorer performance for the normal-hearing group on spectral-ripple detection and discrimination tasks. The results suggest that simulating poorer spectral resolution alone is not sufficient to explain implant user performance. Spectral-ripple detection performance likely reflects both spectral- and intensity-resolution, but few vocoder simulations utilize implant-like intensity resolution. In this study, a vocoder processing strategy simulating both electric field spread and dynamic range limitations was used to evaluate performance of normal-hearing and cochlear-implant listeners on speech-in-noise and spectral-resolution tasks. Data were collected using a novel online remote testing platform developed implemented with the MATLAB Web App Server. A comparison of results between groups highlights the implications of including psychophysically measured estimates of cochlear-implant performance in acoustic simulations of electric hearing. [Work supported by NIH grant R01DC012262.]