ACOUSTIC AND LANGUAGE PROCESSING TECHNOLOGY FOR SPEECH RECOGNITION

This paper describes acoustic and language processing technology for a utomatic speech recognition. Speech recognition systems usually consis t of acoustic and language processing modules. The acoustic processing extracts feature parameter vectors from the speech utterance and perf orms pattern recognition by comparing the vector sequence and pre-defi ned acoustic models. The most likely model is then chosen as the recog nition result. The language processing helps recognition by narrowing down the number of candidates or selects the most linguistically match ing hypothesis from those produced by the acoustic processing.