This paper addresses the problem of speech recognition with signals corrupted by white Gaussian additive noise at moderate SNR. The energy of the noise is not required. A technique based on a lateral inhibition process approximation with multilayer neural nets, and SNR weighting in acoustic pattern matching algorithms is proposed. At the recognition procedure, the local SNR is estimated by means of the autocorrelation function and is taken into account as a weight in a pattern matching algorithm. A general criterion based on weighting the frame influence in decisions according to local SNR is suggested, and modified versions of both HMM arid DTW algorithms have been designed.
Bibliographic reference. Yoma, Nestor Becerra / McInnes, Fergus R. / Jack, Mervyn A. (1995): "Improved algorithms for speech recognition in noise using lateral inhibition and SNR weighting", In EUROSPEECH-1995, 461-464.