SPEECH RECOGNITION SYSTEM WITH MODEL VARIABILITY PRONUNCIATION

摘要

abstract; speech recognition system with a pronunciation variation model relates to the field of speech recognition and is designed to solve problems of processing and recognition of speech. Summary of the utility model is that the speech recognition system, which includes blocks of acoustic models, language models, the pronunciation dictionary, input and signal pre-processing, the parameterization of the speech signal, the speech detection and decoding spoken language in the list of best hypotheses as there are blocks of variation detector pronunciation shaper combined acoustic models for variativnyh sounds and the conversion probabilities of hypotheses using combined acoustic dressed. The achieved technical result is a reduction in the number of errors in the automatic decoding, pattern recognition and processing of natural speech spoken by simulating amendments variability pronunciations. 1 yl.