Expediting the Identification of Impaired Cochlear Implant Acoustic Model Channels Through Confusion Matrix Analysis

Psychophysical studies with cochlear implant patients have demonstrated performance variability across the electrode array, in some cases with electrodes displaying quite anomalous behavior. Aberrant cochlear implant channels have been linked to reduced speech recognition ability and poor transmission of auditory information. However, psychophysical tests to completely assess all psychophysical metrics and detect impaired electrodes are time consuming. This study proposes using the confusion matrices from closed response set listening tests to identify impaired electrodes. A listening experiment was performed in which normal-hearing subjects identified vowel and consonant tokens processed with an impaired cochlear implant acoustic model. Two methods of analyzing the vowel and consonant confusion matrices to identify the impaired acoustic model channels are considered. The performance of the two analysis methods motivates further study of the uses of relatively simple listening tests to provide information that can expedite larger scale testing of psychophysical metrics and impaired electrodes