Speaker Gender Recognition Using Hidden Markov Model

Abstract

Gender is an important demographic attribute of people. With the evolution in modern technologies in various fields of life and entering the computer systems in all applications, this led to the use of transactions instead of these technologies and human speech processing, and speaker recognition technology race.

In this research we build a system to distinguish the gender of the speaker, and through the audio information that has been obtained from the speech signal, passes the system in four phases, namely the phase of initial processing, and phase  of features extraction, we use (MFCC) (Mel Frequency Cepstral Coefficients) technique, then comes the phase of training the EM algorithm was used to achieve the greatest expected limit, and finally the testing phase, which has been applied hidden Markov models in it. All algorithms and programs have been written using the language of Matlab.

Keywords: Gender Recognition, Hidden Markov Model, Mel Frequency Cepstral Coefficients, Speech Recognition