CODING OF SOUND IN A COCHLEAR PROSTHESIS: SOME THEORETICAL AND PRACTICAL CONSIDERATIONS a

The optimal strategy for coding speech as intelligible across an intracochlear multielectrode array reflects a compromise between what should be done and what can be done. That is, it reflects a compromise between simulation of the critical aspects of the representation of speech elements across the auditory nerve array and the idealization of that representation, given practical limitations of the spatiotemporal control of nerve array stimulation with applicable intracochlear electrodes. Recent studies on the auditory nerve array representation of speech elements (especially the studies of Sachs, Young and colleagues coupled with parametric studies with cochlear implant devices ‘k--B indicate, on a first order, what is essential to represent in the distributed excitation of the nerve array. Further, they reveal how high levels of speech recognition are accounted for in the distributed patterns of activation of the auditory nerve array in normalhearing subjects driven with low-channel-number v ~ c o d e r s . ~ These studies suggest that the “optimal” simulation model is a special kind of channel vocoder, especially given limitations of electrical stimulation control of the auditory nerve array. In this report, it will be argued that a distributed nerve-stimulation vocoder model constitutes a rational starting point for progressive development of a multichannel cochlear prosthesis. Nerve array stimulation control required to implement such a model is described. Some basic test results in a patient stimulated with threeand four-channel versions of such a model are described. Physical and practical limitations that impede development of more complex stimulation models are discussed. Finally, strategies for improving the simulation (reducing distortion) is such devices are briefly summarized.