Abstract
Consonant confusion study was performed in a patient with a Nucleus multiple electrode Cochlear implant.
Discrimination between voiced and unvoiced was successful, but conausions occured in voiced plosives, voiced fricatives, nasals, and unvoiced plosives, respectively.
Discharges of the cochlear implant electrodes for the japanese vowel-consonant-vowel sillables were digitally recorded. And discharge patterns for consonants were compared with their sound spectrograms.
Discharges of electrode #20 were always observed in voiced consonants, but not in unveiced. They were thought to correspond to buzz bars in spectrograms. Recognition of voiced or unvoiced may be helped by discharges of electrode #20.
In plosives, the confusions may be caused by poor electrode discharges at the plosive portion of spectrograms. In some voiced consonants, discharges were seen which did not have the corresponding components in their original spectrograms.
In order to make the consonant discrimination better, a new speech processing strategy should be developed which have more fidelity to temporal and frequencial properties of consonants.
