Abstract
In cochlear implants, limited spectral and temporal information is provided. Previous studies argued for different effects of temporal information on speech identification in adverse environments. Particularly, it is unclear how speech intelligibility is influenced by the low-pass cutoff frequency of temporal envelope extractors in noise. The current study explored this issue with Mandarin noise-vocoder simulation in babble noise. Mandarin spoken sentences were mixed with multi-talker babble at different SNRs (4-, 6- And 8-dB) and then noise-vocoded. The vocoders had three numbers of logarithmic-spaced frequency channels (6, 8 and 10) with lowpass filtering the Hilbert envelope at 50 or 500 Hz in each channel. We measured the sentence intelligibility with native normal-hearing subjects and found that intelligibility for the high frequency cutoff condition (500-Hz) is significantly higher at 8-dB SNR with 6 and 8 channels, but significantly lower at 4-dB SNR with 6 channels than the low frequency cutoff condition (50-Hz). This finding suggests that high cutoff frequency of the envelope extractor in noisevocoders can either improve or impair speech intelligibility in babble noise, hinging on the number of channels and the SNR level. Potential implications of the current finding for cochlear implant designs were further discussed in the paper.
Original language | English |
---|---|
Pages (from-to) | 3646-3650 |
Number of pages | 5 |
Journal | Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH |
Publication status | Published - 1 Jan 2013 |
Externally published | Yes |
Event | 14th Annual Conference of the International Speech Communication Association, INTERSPEECH 2013 - Lyon, France Duration: 25 Aug 2013 → 29 Aug 2013 |
Keywords
- Babble noise
- Envelope cutoff frequency
- Intelligibility
- Noise-vocoder
ASJC Scopus subject areas
- Language and Linguistics
- Human-Computer Interaction
- Signal Processing
- Software
- Modelling and Simulation