Disordered voice classification using SVM and feature selection using GA

Abstract

Many individuals are subjected to the risk of voice disorders which may be characterized by hoarseness, vocal fatigue,periodic loss of voice or inappropriate pitch or loudness. These disordered voice cause changes in the acoustic characteristics. Therefore, the voice signal is used as an important measure to diagnose them. This paper deals the classification of normal and two disordered voices using support vector machine (SVM). For this classification the voice signal which is recorded from the patients is used. The mel frequency cepstral coefficients (MFCC), delta and double delta coefficients are extracted as features from the voice signal. For best feature selection, genetic algorithm (GA) is used. The performance of the classifier is enhanced after applying GA.