Wavelet Transforms in Speech Processing

Abstract

This is an era of computers. Computers have become an integral part of human life as they are influencing our lives in every possible way, and are used by every person irrespective of age and sex. The ease with which we can exchange information between user and computer is of immense importance today. But the input devices like mouse and keyboard when used as an interface to exchange the information have some limitations also. Speech which is a natural and quick way of exchanging the information between humans, if used to communicate with computers can overcome all these limitations. Speech recognition is an area of research which has attracted many researchers across the world. Speech is a time varying signal. The information contained in the signal is very difficult to analyze. Traditional methods of speech recognition use Mel Frequency Cepstral Coefficients (MFCC) and Short Time Fourier Transform (STFT) to extract the features out of a speech signal, and the model has been successfully implemented in many speech recognition engines. The wavelet transform has better time-frequency localization property as compared to the Short Time Fourier Transform (STFT). For better feature extraction of the speech signal we use the applications of wavelets, that is, speech recognition. Here, in the present work, the concept of wavelet transform has been used for feature extraction from the speech signals. The speech signal of short duration has been taken for the purpose. All the speech signals are phonemes (fundamental unit of speech). The processed speech signal was fed to a neural network for classification to measure the performance of the feature extraction. For the classification of speech signal vector, the self-organizing feature map (SOFM) has been used which is a type of the unsupervised learning method. The results are good and require further experimentation in future.