To Improve Speech Recognition in Noise for Cochlear Implant Users

Abstract

The hypothesis that when listening to speech in fluctuating maskers, CI users can not fuse the pieces of the message over temporal gaps because they are not able to perceive reliably the acoustic landmarks introduced by obstruent consonants (e.g., stops). To test this hypothesis, CI users were first presented with sentences containing clean obstruent segments, but corrupted sonorant segments (e.g., vowels). The other experiment investigated the hypothesis that envelope compression smears acoustic landmarks which signify syllable/word boundaries. To test this, CI users were presented with noise-corrupted stimuli processed using logarithmic compression during voiced segments and a weakly-compressive mapping function during unvoiced segments. All patients were profound-totally deaf, adults with a post lingual onset of impairment. The data support the efficacy of a feature extraction coding system where specific formant and amplitude information are transmitted via direct electrical stimulation to the cochlea. To examine the hypothesis that the newer generations of cochlear implants could provide considerable speech understanding to late-implanted, prelingually deaf adult patients. I. The Clinical Program 1.1Experiment 1: Masking Releases for Cochlear Implant Users Methods Subjects A total of seven postlingually deafened Clarion CII implant users participated in this experiment. All subjects had at least 3 years of experience with their implant devices. Most subjects visited our lab two times. The biographical data for each subject are given in Table 1.1. Stimuli The speech material consisted of sentences taken from the IEEE database . All sentences were produced by a male speaker. The sentences were recorded in a sound-proof booth in our lab at a 25 kHz sampling rate. Two types of maskers were used. The first was speech-shaped noise, which is continuous (steady- state) and had the same long-term spectrum as the test sentences in the IEEE corpus. Table 1.1. Biographical data of the CI users