Union: A new approach for combining sub-band observations for noisy speechrecognition

Recent studies have shown that the sub-band based speech recognition approa ch has the potential of improving upon the conventional, full-band based mo del against frequency-selective noise. A critical issue towards exploiting this potential is the choice of the method for combining the sub-band obser vations. This paper introduces a new method, namely, the probabilistic-unio n model, for this combination. The new model is based on the probability th eory for the union of random events, and represents a new method for modeli ng partially corrupted observations given little knowledge about the corrup tion. The new model has been incorporated into a hidden Markov model (HMM) and tested for recognizing a speaker-independent E-set, corrupted by variou s types of additive noise. The results show that the new model offers robus tness to partial frequency corruption, requiring little or no knowledge abo ut the noise statistics. (C) 2001 Elsevier Science B.V. All rights reserved .