Voice Disorder Signal Classification Using M-Band Wavelets and Support Vector Machine

摘要

The aim of this work is to present a noninvasive method to classify normal voice signals and those corresponding to voice disorders. Use of wavelet decomposition is prevalent for feature extraction in this field and provides good frequency resolution in lower-frequency subbands. In this work, to provide better frequency resolution in higher-frequency subbands as well, we use M-band wavelet decomposition for feature extraction, employing a genetic algorithm to determine the parameters of the optimal wavelet. Moreover, a support vector machine is used as the final classifier. By employing a well-known pathological voice database, normal and pathological cases are classified using a five-band wavelet system for feature extraction, showing good performance.