Cochlear and cognitive models for speech enhancement
| Autor: | Andrew Dittberner, Jeff Bondy, Melissa Dominguez |
|---|---|
| Rok vydání: | 2006 |
| Předmět: |
Masking (art)
Acoustics and Ultrasonics Computer science Noise reduction Acoustics Speech recognition Cognition Intelligibility (communication) computer.software_genre Speech enhancement Arts and Humanities (miscellaneous) otorhinolaryngologic diseases Dynamic range compression Audio signal processing computer Cognitive load |
| Zdroj: | The Journal of the Acoustical Society of America. 120:3129-3130 |
| ISSN: | 0001-4966 |
| DOI: | 10.1121/1.4787703 |
| Popis: | There has been a long running interest in using cochlear models for audio processing. Psychophysical masking stemming from normal cochlear operation is used as the basis for most audio compression circuits, while the cochleas mel‐frequency response is mimicked by most automatic speech recognition circuits. An area that has been less well explored in audio processing is the auditory brain. Recent research shows that cognitive abilities are a good predictor of peformance on many hearing tasks, especially those which are more complex and realistic. By extending a nonlinear, adaptive cochlear model with cognitive modules a novel tool for sound assessment is made that allows an estimate of cognitive load. Quantitative offline benefit has never been modeled before. We test the new predictor on noise reduction circuits in hearing aids. This family of algorithms is well known to not increase intelligibility but has been shown to reduce perceived noisiness. This reduction is proposed as equivalent to reducing cognitive load. The elderly, with their heavier reliance on high level processing, could greatly benefit from the reduction in cognitive load. A new algorithm for reducing cognitive load is compared to spectral subtraction algorithms. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|