Speech Command Recognition by using Dynamic Time Warping

Abstract

Speech Recognition is a technology that can be useful in many applicationof our daily life. This proposed method present the speaker dependent speech recognition system. In this paperinclude the three steps of speech command recognition systems. Firstly, automatically crop the input command signal for removing silent part of the command .The second one is extract the feature vectors by using Mel frequency cepstralcoefficients (MFCCs) and then feature matching the command and predefined speech command by using Dynamic Time Warping (DTW). Before matching with DTW, the system sorts the predefined command for matching with command to improve performance of recognition. This system uses the ten predefined command for one username. This paper also compares the speech recognition system between the using of Multi-band Spectral Subtraction and Feature extractions in preprocessing method. Finally, the proposed system designs the “Information Retrieval” application using MATLAB.