Evolutionist approach and MFCC modeling for Arabic automatic recognition

Abstract

In this article, we suggest a system for automatic recognition of isolated Arabic words, it is a multi-speaker system, even independent of speaker and robust in a noisy environment, it uses a genetic algorithm for recognition, and the Mel frequency cepstral coefficients (MFCC) to modelise the speech signal, it was implemented with the matlab7 platform language. A new mutation method (injection mutation) is proposed and used in the genetic algorithm. To evaluate the performance of the system, we have made an oral corpus that represents the most Arabic language characteristics, it could be used by other researchers to test and validate their systems working on Arabic language.

Keywords: Automatic speech recognition, genetic algorithm, Arabic language, Mel frequency cepstral coefficients (MFCC), language corpora.