Text-dependent speaker verification using genetic algorithm and competitive learning neural network

Recently, biometric systems are of interest to researchers in many areas. Biometric systems are automated methods of verifying or recognizing the identity of a living person on the basis of some physiological characteristic, like fingerprint or iris pattern, or some aspects of behavior, like handwriting or key-stroke patterns[1]. In this paper, we present a text-dependent speaker verification using behavioral characteristics of speaker's voice signal, competitive learning neural networks and genetic algorithm. For speaker verification we acquire the speech from a user and digitize it at 11kHz. From the digitized samples, LPC-Cepstrum coefficients are computed and used as features. Simple competitive leaming(SCL) neural networks are learned using these features. Genetic algorithm searches the optimal threshold value for verifying a speaker. Speech pattern of the test speaker is compared with the stored reference patterns of neural networks, If it is within the prescribed range(threshold value), the speaker is authorized. Experimental results indicate that speaker verification can be used for security entry and access control.