Speech Recognition with Temporal Envelope and Fine Structure Cues Under Simulated Reverberation

*, Arivudai Nambi. P and Bhat, Jayashree. S. and *, Rajesh Ranjan (2010) Speech Recognition with Temporal Envelope and Fine Structure Cues Under Simulated Reverberation. International Journal of Computational Intelligence and Healthcare Informatics, 3 (1). pp. 41-44.

PDF 7_Arivudai_Nambi,_P,_Jayashree._S._Bhat_&_Rajesh_Ranjan.pdf - Published Version Restricted to Registered users only Download (283kB) | Request a copy

Abstract

Auditory listening devices that use only envelope cues for speech recognition perform poorly under many adverse listening conditions. Current study investigated speech recognition with (i) only envelope and (ii) envelope + fine structure under simulated reverberation. Envelopes were extracted from 8 bands at 160Hz cutoff and fine structure was extracted at 300Hz cutoff. Reverberation time of 1200msec was used. 30 normal hearing adults listened to the synthesized sentences. Sentence recognition scores were poor only with envelope cues but significantly improved with the addition of fine structure. Current study suggests that speech coding strategies should consider fine structure cues to improve the performance under reverberation.

Actions (login required)

View Item