Clear speech acoustics and perception: Implications for hearing aid design.

摘要

Previous studies showed that clear speech had higher intelligibility than conversational speech. Here we studied temporal acoustic cues that contribute to the clear speech advantage and attempted to digitally manipulate them to enhance speech intelligibility for the hearing impaired, particularly those with auditory neuropathy and cochlear implants. We conducted speech perception experiments over a wide range of signal-to-noise ratios in normal-hearing, cochlear-implant, and auditory-neuropathy subjects. We found that clear speech produced higher intelligibility than conversational speech in all subject populations. We used cochlear-implant simulation, "auditory chimera", and combined low- and high-pass filtering techniques to investigate acoustic cues that are responsible for the observed clear speech advantage. We found that intrinsic acoustic properties other than the speaking rate contribute the most to the clear speech advantage. The intrinsic acoustic cues may include both temporal envelope and temporal fine structure and are distributed across a wide range of frequencies. Moreover, we found that the temporal envelope contribute more to the clear speech advantage at high signal-to-noise ratios, the temporal fine structure contribute more to the clear speech advantage at low signal-to-noise ratios. Based on the knowledge gained from psychophysics as well as acoustic analysis and perception of clear speech, we proposed several novel speech processing strategies to improve speech recognition in auditory-neuropathy subjects. The novel strategies included (1) enhanced temporal modulation that compensates for their impaired temporal processing ability, (2) frequency high-pass and upward-shifting strategies that minimize the impaired pitch perception at low frequencies. We provided preliminary evidence for improved performance by the high-pass processing scheme in auditory-neuropathy subjects. The frequency and formant upward shifting strategies altered the voice quality and/or the normal tonotopic map, requiring further long-term studies to demonstrate their effectiveness.