ISESCO JOURNAL of Science and Technology
The Official Journal of ISESCO Centre for Promotion of Scientific Research

Abstract

This article presents an automatic voice recognition system of isolated words using Support Vector Machines (SVM). Three phases are necessary to achieve this system: the first one involves the extraction of the vocal signal’s parameters, and the second one the training and regulation of the SVM model’s parameters to improve their performance in recognition. The last identification phase allows the classification of the words available in the NTIMIT telephonic database. The results obtained show that the use of SVM is a very effective and promising technique.

Keywords: Automatic system of isolated word recognition, Acoustic analysis, Regulation of parameters, Statistical learning, SVM model.