Speech recognition with altered spectral distribution of envelope cues

Recognition of consonants, vowels, and sentences was measured in conditions of reduced and altered spectral information. Speech materials were passed through four bandpass filters (analysis bands). The output of each band was half-wave rectified, and low-pass filtered at 160 Hz to extract the temporal envelope. The envelope from each band modulated a bandlimited noise (carrier bands). Experiment I explored the importance of the cutoff frequencies of the bands, which proved not to be a critical parameter for speech recognition as long as the analysis and carrier bands were the same. Experiment II warped the spectral distribution of envelope cues by mismatching the analysis and carrier bands in frequency extent, resulting in a dramatic decrease in performance. Experiment III simulated the spectral smearing caused by broad auditory filters or electrode interaction in a cochlear implant by decreasing the slopes on the noise carrier bands. A decrease in performance was observed only for the most broadly overlapping carrier bands, indicating that speech pattern recognition is robust to spectral smearing. Overall, these results show that, for four bands, the most critical parameter is the frequency alignment of the analysis bands and carrier bands. Supported by NIDCD.