Comparison between Different Methods used in MFCC for Speaker Recognition System

Abstract

The idea of the Speaker Recognition Project is to implement a recognizer which might determine an individual by process his/her voice. The essential goal of the project is to acknowledge and classify the speeches of various persons. This classification is especially supported extracting many key options like Mel Frequency Cepstral Coefficients (MFCC) from the speech signals of these persons by mistreatment methodology of feature extraction method. The on top of options could encompass pitch, amplitude, frequency etc. employing an applied math model like gaussian mixture model (GMM) and options extracted from those speech signals we have a tendency to build a novel identity for every one that listed for speaker recognition. Estimation and Maximization formula is employed, a chic and powerful methodology for locating the most chance answer for a model with latent variables, to check the later speakers against the information of all speakers who listed within the information. Use of divisional Fourier rework for feature extraction is additionally recommended to enhance the speaker recognition potency.