OUTPUT PROBABILITY COMPUTATION CONTROL METHOD OF CONTINUOUSLY DISTRIBUTED TYPE HIDDEN MARKOV MODEL(HMM), TABLE GENERATING METHOD AND VOICE RECOGNITION DEVICE

摘要

PROBLEM TO BE SOLVED: To simplify the output probability computations of the HMM in voice recognition. ;SOLUTION: A first table 410 makes the classification of observed feature vectors. A second table 412 gives specific constants to individual parameters of the feature vectors for every classification given by the table 410. Employing the tables 410 and 412 and the information of the observed feature vectors, the attributed class of the feature vectors is decided 402 while referring to the table 410, the constants corresponding to the parameters are obtained 403 for every class of the obtained feature vectors while referring to the table 412, an output probability is estimated by using the obtained constants in a prescribed inference equation 405 and the output probability of a continuously distributed HMM in voice recognition is computation controlled.;COPYRIGHT: (C)1998,JPO