Speech Recognition Methods for the Robust Distant-talking Speech Recognition System

摘要

A speech recognition method for a robust remote speech recognition system is provided to apply a right and left phoneme connection method and a noise synthesis and adaptation algorithm to pattern matching, thereby improving the recognition rate of the remote speech recognition system. A speech recognition method for a robust remote speech recognition system comprises the following steps of: receiving an input speech, generating a pronunciation queue through a phoneme-unit pronunciation queue generator, converting the pronunciation queue into a triphone in a recognition unit, and recognizing the converted triphone(S61,S62,S63); combining the end phoneme of a word and the first phoneme of a next word to generate a triphone queue which does not appear in the speech recognition in recognizing the triphone; and recognizing the speech for the continuous input word using the recognized triphone(S65).