Noisy speech recognition based on HMMs, Wiener filters and re-evaluation of most likely candidates

The effects of noise on observation probability are described in terms of the mean and variance of the disturbance. The observation that in many cases of erroneous speech recognition the correct model is among the few with the highest scores forms the basis for filtering and reevaluation of a few high scoring candidates before making a final decision. The noisy input signal is filtered by state-dependent Wiener filters derived from the most likely state sequence of each HMM (hidden Markov model). A revised score for each candidate model and the filtered signal is calculated. The method based on state-dependent Wiener filters and reevaluation of highest scoring candidates results in substantial improvement in recognition performance. The increase in computational complexity is not prohibitive and depends on the number of candidates reevaluated.<>