Automatic speaker recognition using a unique personal feature vector and Gaussian Mixture Models

This article presents an automatic speaker recognition system implemented in Matlab, which uses a unique feature vector, the so-called “Voice Print” (VP), to describe the voice. The system uses Gaussian Mixtures Models (GMM) in the classification process. The final part of the paper presents research on the efficiency of speaker recognition for different variants of the system, as well as the results of optimisation of the system.