Method for compensating probability density function, method and apparatus for speech recognition thereby

摘要

Hidden Markov probability density function modelreul compensation method for speech recognition method and apparatus is disclosed accordingly . Probability density function compensation method of the present invention comprises the steps of learning the model having a plurality of probability density function is the probability to recognize the voice signal , when extracting feature vectors from the speech signal, and input the feature vector to increase ; When the learning is complete , obtaining a global distributed by averaging the distribution of each of the probability density function ; Calculating a compensation factor using a global distribution ; And applying each of the probability density function and the global variance , using the compensation factor characterized in that it comprises the step of compensating each of the probability density function by compensating for the global distribution .