Influence of Consonant-Vowel Intensity Ratio on Speech Perception for Hearing Impaired Listeners

Abstract

Speech recognition in noise is an extremely challenging task, particularly for hearing-impaired listeners. One way to alleviate this difficulty is to speak 'clearly' as opposed to 'conversationally' to these individuals. Preprocessing speech with relevent acoustic modifications is expected to improve speech intelligibility for impaired listeners [1, 2, 3, 4, 7 and 13]. The present study seeks to determine the effect of one such clear speech attribute called 'consonant-vowel intensity ratio' on the improvement of speech intelligibility for subjects with sensorineural hearing loss. Perception experiments were conducted using preprocessed stimuli consisting of synthesized plosives of English language on five listners in various levels of comb-filtered additive white noise. The 'consonant recognition scores' and 'Relative information transmitted' for several consonant features were calculated from the obtained confusion matrices. Analysis of confusion matrices indicated significant intelligibility improvement for voiced, unvoiced stops, under labial, alveolar and velar placing feature. The results show improvement of consonant recognition scores, and the relative information transmitted under both the vowel-dependent and vowel-independent cases. The maximum rise in percent correct was 18% points (RTR) and 19% points (RS) at 12 dB SNR.