Method of the textdependent voice authentication

Abstract

In this work the authors propose method of user’s biometric authentication by parameters of his voice. The principal parameters of the developed model are individual speed of the speech and pronouncing length of the speech elements. In the cource of voice samples comparison algorithm development it was discovered that the analysis of the speech and pronouncing length of the speech elements in text dependent user’s voice authentication procedure is reduced to finding the percentage of recording segments, overlapping in content and location, with due consideration for the variability of the uttered passphrase total length. For comparison the Mel-frequency cepstral coefficients method was used.