Combination of feature weight and speech enhancement

In this paper we propose a novel algorithm for feature weight and speech enhancement, in which every feature is weighted according to their credible probability, especially, the weight factors are formulated and obtained from the gain coefficients generated as a by-product of speech enhancement based on short-time spectrum amplitude (STSA) estimation. Moreover, this algorithm can be combined with the front-end speech enhancement itself. Experiments demonstrate that this algorithm affords significant performance improvements at low SNRs.

[1]  Phil D. Green,et al.  Robust automatic speech recognition with missing and unreliable acoustic data , 2001, Speech Commun..

[2]  Xu Yifang,et al.  Robust recognition of noisy speech using speech enhancement , 2000, WCC 2000 - ICSP 2000. 2000 5th International Conference on Signal Processing Proceedings. 16th World Computer Congress 2000.

[3]  Jon Barker,et al.  Soft decisions in missing data techniques for robust automatic speech recognition , 2000, INTERSPEECH.

[4]  David Malah,et al.  Speech enhancement using a minimum mean-square error log-spectral amplitude estimator , 1984, IEEE Trans. Acoust. Speech Signal Process..

[5]  A.V. Oppenheim,et al.  Enhancement and bandwidth compression of noisy speech , 1979, Proceedings of the IEEE.