Design and impelmentation of a speaker verification system using i-vector and Support Vector Machines

Abstract

In the recent years, design of robust and effective speaker verification algorithms has attracted significant research effort from academic and commercial institutions. Speaker verification is a procedure that uses voice properties to prove the identity of person. Speaker verification solutions offer a highly accurate and efficient method of authenticating a person's identity by analyzing his/her voice. It is widely believed that speaker verification systems perform better when there is sufficient background training data to deal with nuisance effects of transmission channels. For some applications however, training data from the same type of sound environment is scarce, whereas a considerable amount of data from a different type of environment is available. This paper, presents a new algorithm for text-independent speaker verification systems based on the extraction of parameters (i-vectors) model. Support Vector Machines is used as a classifier and the model is trained and tested using data of 100 human voices. Evaluation of the data shows acceptable performance of the implemented algorithm. Also implementation of this algorithm using HM 2007 board is part this work. All designed circuits for the latter are presented and results of the trained algorithm mentioned.