METHOD AND APPARATUS FOR SPEECH END-POINT DETECTION USING ACOUSTIC AND LANGUAGE MODELING KNOWLEDGE FOR ROBUST SPEECH RECOGNITION

摘要

A method and apparatus for detecting a voice endpoint using acoustic and language modeling information for robust voice recognition are presented. The method for detecting a voice endpoint according to an embodiment includes a Recurrent Neural Network (RNN)-based acoustic embedding extractor, a phoneme embedding extractor, and a decoder embedding extractor that inputs an acoustic feature vector sequence extracted from a microphone input signal. step; constructing a feature vector by combining sound embedding, phoneme embedding, and decoder embedding in the sound embedding extractor, the phoneme embedding extractor, and the decoder embedding extractor; and inputting the combined feature vector into a deep neural network (DNN)-based classifier to detect a voice endpoint.