Related Experiment Videos
Evaluation of two voice-separation algorithms using normal-hearing and hearing-impaired listeners
1MRC Institute of Hearing Research, University Park, Nottingham, United Kingdom.
The Journal of the Acoustical Society of America
|October 1, 1988
Summary
Two signal-processing algorithms effectively separate simultaneous speech by exploiting voice harmonic structure. Harmonic selection offers superior voice separation, showing potential for hearing aid noise reduction.
Area of Science:
- Speech processing
- Acoustics
- Signal processing
Background:
- Simultaneous speech in a single channel presents a challenge for intelligibility.
- Existing algorithms struggle to separate voices with similar intensities.
Purpose of the Study:
- To compare two signal-processing algorithms for separating simultaneous voiced speech.
- To evaluate the effectiveness of these algorithms for normal-hearing and hearing-impaired listeners.
Main Methods:
- Two algorithms were developed: cepstral filtering and harmonic selection.
- Perceptual evaluations involved vowel separation and consonant-vowel word recovery.
- Tests included normal-hearing and hearing-impaired listeners.
Main Results:
- Both algorithms improved speech identification accuracy for all listeners.
- Intelligibility enhancement was observed across various fundamental frequency (F0) separations.
- Harmonic selection demonstrated more effective voice separation than cepstral filtering.
Conclusions:
- Signal-processing algorithms can enhance speech intelligibility in multi-talker scenarios.
- Harmonic selection shows promise for noise reduction in digital hearing aids.
- Further research could integrate these algorithms into assistive listening devices.