Auditory evoked responses to perceived quality of vocoded speech

Abstract

One of the common uses of vocoded speech is to simulate what speech would sound like when it is processed via a cochlear implant. However, listeners with normal hearing may perceive vocoded speech differently depending on the carrier signals (noise or tone) or channel bandwidths. In this study, we attempted to determine how physiological measures correlate with the quality ratings of vocoded speech that varied in carrier signal and channel bandwidth. Eleven NH subjects rated the perceived sound quality of a spoken sentence processed with noise and tone vocoders (22-channel analysis/synthesis filterbank) whose channels were broadened and narrowed by different Q factors. Each subject was instructed to rate sound quality on a 10 point scale where “10” indicates “clean” and “1” represents “very-distorted”. Quality ratings for six different combinations of carrier and Q factor were very consistent across listeners. Auditory evoked potentials (MMN) were obtained from three NH listeners. Vowel stimuli processed through the vocoders with variations in the Q value were used as stimuli. Changes in N1 latency were observed to correlate with the perceived quality ratings of the vocoded speech. [This study was supported by NIH-K25-DC010834 (PI: Tan), PSC-CUNY (PI: Martin), NIH-R01-DC011329 (PI: Neuman and Svirsky), and NIH-R01-DC003937 (PI: Svirsky).]