A comparison of the speech understanding provided by acoustic models of fixed-channel and channel-picking signal processors for cochlear implants.

Vowels, consonants, and sentences were processed by two cochlear-implant signal-processing strategies-a fixed-channel strategy and a channel-picking strategy-and the resulting signals were presented to listeners with normal hearing for identification. At issue was the number of channels of stimulation needed in each strategy to achieve an equivalent level of speech recognition in quiet and in noise. In quiet, 8 fixed channels allowed a performance maximum for the most difficult stimulus material. A similar level of performance was reached with a 6-of-20 channel-picking strategy. In noise, 10 fixed channels allowed a performance maximum for the most difficult stimulus material. A similar level of performance was reached with a 9-of-20 strategy. Both strategies are capable of providing a very high level of speech recognition. Choosing between the two strategies may, ultimately, depend on issues that are independent of speech recognition-such as ease of device programming.