Release of masking and FAME performance evaluation to improve speech intelligibility on cochlear implant

Abstract

Cochlear implant (CI) is a hearing aid for people with profound deafness, inability to respond to a sound stimulus above 90dB SPL. The main problem of CI user is the inability to discriminate simultaneous incoming sounds, focusing on the desired sound (target) whilst ignoring the rest (cocktail party problem). In this research, the release of masking strategy is introduced to give a glimpse of acoustical cues for CI user to make the discrimination easier in the presence of background noise (masker). We evaluated the performance release of masking with various combinations of target sound channel vocoder (the sound that want to hear) with respect to masker channel. For further enhancement, we also evaluated spectral subtraction (SS), Ephraim-Malah, and Frequency Amplitude Modulation Encoding (FAME) to improve intelligibility of channelized sounds. The results showed that the release of masking may have improve the intelligibility on CI users to access temporal structure of waveform of simultaneous stimuli. The best enhancement method is FAME with PESQ value 1.4 and % correct word increased to 72%.