Signal coding in cochlear implants: exploiting stochastic effects of electrical stimulation.

Abstract

Speech perception in quiet with cochlear implants has increased substantially over the past 17 years. If current trends continue, average monosyllabic word scores will be nearly 80% by 2010. These improvements are due to enhancements in speech processing strategies, to the implantation of patients with more residual hearing and shorter durations of deafness, and to unknown causes. Despite these improvements, speech perception in noise and music perception are still poor in most implant patients. These deficits may be partly due to poor representation of temporal fine structure by current speech processing strategies. It may be possible to improve both this representation and the dynamic range of electrical stimulation through the exploitation of stochastic effects produced by high-rate (eg, 5-kilopulse-per-second) pulse trains. Both the loudness growth and the dynamic range of low-frequency sinusoids have been enhanced via this technique. A laboratory speech processor using this strategy is under development. Although the clinical programming for such an algorithm is likely to be complex, some guidelines for the psychophysical and electrophysiological techniques necessary can be described now.