Speech Recognition using Wavelet based Feature Extraction Techniques

Abstract

Speech recognition by machine may be defined as the conversion of human speech signal into textual form automatically by the machine without any human intervention. Two feature extraction techniques utilizing DWT (Discrete Wavelet Transform) and WPD (Wavelet Packet Decomposition) for speech recognition are discussed in the present article. The comparison of two speech recognizer, first, based on Discrete Wavelet Transform and the second based on Wavelet Packet Decomposition, and with four classifiers has been done in this paper. The proposed method is implemented for a database consisting of ten digits and two hundred speakers, making it a database of 2000 speech samples. The results present the accuracy rate of the respective speech recognizers.