An Automatic Speaker Recognition System

Abstract

Automatic speaker recognition is the use of a machine to recognize a person from a spoken phrase i.e. it is the process of automatically recognizing the person who is speaking on the basis of information included in an individual’s speech. It is one of the applications of non-invasive Bio-metric recognition. This speaker recognition technique makes it possible to use a person’s voice to verify their identity and control access to services such as voice dialing, banking by telephone, telephone shopping, database access services, information services, voice mail, security control for confidential information areas, and remote access to computers. The purpose of this project was to research the existing speaker recognition techniques and to implement one of them for an employment of an automatic speaker recognition system. The project was carried out in the following three phases.  In the first phase, the digital speech data acquisition was performed.  In the second phase, we carried out the prominent feature extraction. We used Mel Frequency cepstrum coefficient (MFCC) method for feature extraction.  In the third phase, pattern recognition for feature matching was applied over extracted features. This involved the clustering of training feature vectors and storing in a speaker data base. Vector Quantization (VQ) method for Feature matching and the modeling of a speaker was used. Decision making through pattern matching was the last step.