Clear and conversation speech perception in simulated cochlear implant and simulated electro acoustic stimulation

Abstract

Perception of speech is an important goal for persons with Cochlear Implant (CI) and Electro Acoustic Stimulation (EAS). It has been argued that naturally produced clear speech is more intelligible to such persons than conversational speech. The factors responsible for clear speech advantage are increased voice intensity slower rate, frequent pauses, greater vowel duration and dynamic formant movements. The relative advantage of clear speech in CI and EAS are not yet well understood. The present study used 20 native Hindi Speaking (age range 20-30 years) participants on a listening task. The stimuli were 90 previously selected Hindi sentences processed with a noise band vocoder implemented in Matlab. The sentences were spoken by a native Hindi speaker which was recorded in both Clear and Conversational speaking styles & mixed with four talker babble at -4 dB SNR. CI processing simulated using 8-channel noise-band vocoder. The sentences were filtered to 8 bands and the envelopes were extracted from each band. The carrier noise bands were modulated by the envelopes and resynthesized to produce the processed speech. Simulation of EAS was achieved with low pass stimulus (630Hz) and upper five channels of the eight-channel vocoder. Verbatim responses of the subjects were recorded and scored for accuracy. The overall results indicated better understanding in clear speaking style than in conversational style, in simulated CI condition as well as simulated EAS condition. The data was consistent with previous studies. Clear speech had an advantage in improving speech perception whenever there werefewer cues for speech perception due to noise in CI and EAS. The details of comparison of CI and EAS conditions will be discussed.