Related Experiment Videos
Dynamic formant tracking of noisy speech using temporal analysis on outputs from a nonlinear cochlear model
1Department of Electrical and Computer Engineering, University of Waterloo, Ont., Canada.
IEEE Transactions on Bio-Medical Engineering
|May 1, 1993
Summary
This study models speech processing in the auditory system, developing algorithms for accurate formant extraction even in noisy conditions. The research demonstrates robust tracking of speech formants using a novel cross-channel correlation method.
Area of Science:
- Auditory Neuroscience
- Speech Processing
- Computational Auditory Modeling
Background:
- Understanding how the peripheral auditory system represents speech in noise is crucial for hearing aid development.
- Existing methods for formant extraction struggle with dynamic speech and high noise levels.
Purpose of the Study:
- To model the temporal responses of the peripheral auditory system to speech and speech in noise.
- To develop and evaluate algorithms for automatic formant extraction from dynamic, noisy speech signals.
Main Methods:
- A basilar membrane model with laterally coupled damping elements was employed.
- A cross-channel correlation algorithm and interpeak interval analysis were devised for formant extraction.
- The model's digital implementation and numerical properties were described.
Main Results:
- Simulation results showed tonotopically organized temporal response patterns related to speech formants.
- Noise levels had minimal influence on these response patterns.
- The cross-channel correlation algorithm accurately tracked formant movements in speech.
Conclusions:
- The developed auditory model effectively represents formant frequencies in speech and speech in noise.
- The cross-channel correlation algorithm provides a robust method for automatic formant extraction.
- This approach holds promise for improving speech intelligibility in noisy environments.