Learning to recognize spectrally reduced speech: A cochlear implant simulation study

Abstract

This study investigated training-related improvement in open-set identification of sentences that were spectrally reduced to four channels using the CIS processing strategy [Shannon et al., J. Acoust. Soc. Am. 104, 2467–2476 (1998)]. Over the course of 11 training blocks, listeners with normal hearing identified Harvard IEEE [IEEE Trans. Audio Electroacoust. 17, 225–246 (1969)] sentences with feedback. Half of the listeners listened to a male talker during training; the other half to a female talker. Listeners were tested at different points with blocks of IEEE sentences spoken by the familiar talker, blocks of IEEE sentences by the novel talker, blocks of IEEE sentences by the two talkers randomly intermixed, and blocks of HINT sentences [Nilsson et al., J. Acoust. Soc. Am. 95, 1085–1099 (1994)]. Preliminary data showed an improvement in sentence identification over the course of training. The data also showed a trend of talker specificity for such improvement, in that sentences produced by the novel speaker were harder to identify. In some instances, random intermixing of talkers led to poorer performance than having just one talker within a testing block. Implications for theories of auditory memory and for practical aspects of training will be discussed. [Work supported by URC, Univ. of Cincinnati.]