A NOVEL PROJECTION-BASED LIKELIHOOD MEASURE FOR NOISY SPEECH RECOGNITION

The projection-based likelihood measure, an effective means of reducin g noise contamination in speech recognition, dynamically searches an o ptimal equalization factor for adapting the cepstral mean vector of hi dden Markov model (HMM) to equalize the noisy observation. In this pip er, we present a novel likelihood measure which extends the adaptation mechanism to the shrinkage of covariance matrix and the adaptation bi as of mean vector. A set of adaptation functions is proposed for obtai ning the compensation factors, Experiments indicate that the likelihoo d measure proposed herein can markedly elevate the recognition accurac y. (C) 1998 Elsevier Science B.V. All rights reserved.