The contribution of amplitude modulations from speech and competing noise sources for speech recognition
Daniel Fogerty · The Journal of the Acoustical Society of America · 2016
Speech is often heard in the presence of background noise during everyday listening conditions. Concurrent amplitude modulations of the speech and the interfering noise result in portions (i.e., “glimpses”) of the speech signal that remain preserved at favorable signal-to-noise ratios while other portions of speech may be highly degraded. The acoustic and phonetic properties preserved during these glimpses play a role in determining the overall sentence intelligibility. Under such conditions, vocalic cues are relatively more preserved than consonantal/obstruent cues due to average differences in intensity. Investigations examining these different acoustic cues suggest important perceptual contributions from preserved speech amplitude modulations and non-simultaneous interactions from competing noise amplitude modulations. Results from several studies suggest that conditions that maximize the preservation of the sentence temporal envelope (e.g., during long glimpses or for vocalic intervals) result in higher levels of speech recognition. Amplitude modulation during noise-dominated intervals also impacts speech recognition performance. Overall, results suggest that the relative preservation and interaction (e.g., modulation masking or perceptual restoration) of amplitude modulations from the speech and noise is crucial for predicting speech recognition during noise interruption. [Work supported by NIH and ASHA.]