Analysis of MFCC and BFCC in a speaker identification system

摘要

The most significant factor of interaction among human being is language and speech is utilized as the medium. A parametric form of a signal is used by the speech recognizers to attain the peak imperative distinct features of communication signal for recognition reasons. Different feature extraction techniques used to extract the distinguishable characteristics of the speech signal. In this paper, the performance of Gaussian Mixture Model (GMM) based Mel-frequency Cepstral Coefficients (MFCC) and bark frequency Cepstral coefficients (BFCC) speaker identification system has been analyzed on the basis of identification rate, number of Speaker, gender and computational time. It is found that the GMM based MFCC is the optimum feature extraction technique as comparing to BFCC.