Performance analysis of Support Vector Machine as classifier for voiced and unvoiced speech

The classification of speech into voiced and unvoiced is often a very important step in speech recognition and speaker identification systems. Even though several methods have been proposed for the same, the classification has been met with only moderately accurate results. In this paper, we propose the use of Support Vector Machines for classifying the speech signals. Support Vector Machines have been highly successful classifiers in general with their hyperplane approach. On the contrary to the usual approach with the use of SVM in its various applications, the focus in this approach is the trimming of the training set. The results obtained are compared with the most commonly used and successful methods for speech classification, the Bayesian classifier and Linear Predictive Coding (LPC).