PERFORMANCE EVALUATION OF STATISTICAL APPROACHES FOR AUTOMATIC TEXT-INDEPENDENT GENDER IDENTIFICATION SYSTEM

In this paper, robust feature for Automatic text-independent Gender Identification System has been explored. Through different experimental studies, it is demonstrated that the timing varying speech related information can be effectively captured using Hidden Markov Models (HMMs) than Gaussian Mixture Models (GMMs) . The study on the effect of feature vector size for good Gender Identification demonstrates that, feature vector size in the range of 18-22 can capture Gender related information effectively for a speech signal sampled at 16 kHz, it is established that the proposed Gender Identification system requires significantly less amount of data during both during training as well as in testing. The Gender Identification study using robust features for different states and different mixtures components, training and test duration has been exploited on TIMIT database.