CONF
astrid-01-05a/IDIAP
Adaptive ML-Weighting in Multi-Band Recombination of Gaussian Mixture ASR
Hagen, Astrid
Bourlard, Hervé
Morris, Andrew
HMM/ANN-Hybrid
ML-adaptation
multi-band
weighting
EXTERNAL
https://publications.idiap.ch/attachments/reports/2001/rr01-05.pdf
PUBLIC
https://publications.idiap.ch/index.php/publications/showcite/astrid-01-05
Related documents
ICASSP
1
257-260
2001
Multi-band speech recognition is powerful in band-limited noise, when the recognizer of the noisy band, which is less reliable, can be given less weight in the recombination process. An accurate decision on which bands can be considered as reliable and which bands are less reliable due to corruption by noise is usually hard to take. In this article, we investigate a maximum-likelihood (ML) approach to adapting the combination weights of a multi-band system. The Gaussian Mixture Model parameters are kept constant, while the combination weights are iteratively updated to maximize the data likelihood. Unsupervised offline and online weights adaptation are compared to use of equal weights, and `cheating' weights where the noisy band is known, as well as to the fullband system. Initial tests show that both ML-weighting strategies show a robustness gain on band-limited noise.
REPORT
astrid-01-05/IDIAP
Adaptive ML-Weighting in Multi-Band Recombination of Gaussian Mixture ASR
Hagen, Astrid
Bourlard, Hervé
Morris, Andrew
HMM/ANN-Hybrid
ML-adaptation
multi-band
weighting
EXTERNAL
https://publications.idiap.ch/attachments/reports/2001/rr01-05.pdf
PUBLIC
Idiap-RR-05-2001
2001
IDIAP
Martigny, Switzerland
February 2001
Published in: ICASSP, Salt Lake City, Utah, USA, May 2001
Multi-band speech recognition is powerful in band-limited noise, when the recognizer of the noisy band, which is less reliable, can be given less weight in the recombination process. An accurate decision on which bands can be considered as reliable and which bands are less reliable due to corruption by noise is usually hard to take. In this article, we investigate a maximum-likelihood (ML) approach to adapting the combination weights of a multi-band system. The Gaussian Mixture Model parameters are kept constant, while the combination weights are iteratively updated to maximize the data likelihood. Unsupervised offline and online weights adaptation are compared to use of equal weights, and `cheating' weights where the noisy band is known, as well as to the fullband system. Initial tests show that both ML-weighting strategies show a robustness gain on band-limited noise.