Development of neural network database for biometric images based on several parameters of voice sound spectrum for authentication and authorization by voice

Abstract

This study focuses on the methods of biometric authentication and identification of network administrators by voice. The authors propose a method for presenting unique voice parameters, as well as the principle of constructing an artificial neural network for a user authentication system and access control. The structure of the biometric authentication / authorization system of information system users is presented using unique voice parameters: the frequency range of the vowel fourth formant, the fourth formant frequency and the leading formant frequency of spectral characteristic. Two neural network databases of biometric images were built: the first is based only on the characteristics of the vowel fourth formant, the second uses the frequency of the leading formant as well. Possible solutions for specific tasks of building and teaching ANNs are considered to build a base of user biometric images and reduce first and second types errors.

Keywords

• biometric image of a person, neural network database of biometric images, characteristics of the fourth formant of the vowel sound, frequency of the leading spectral characteristic formants, voice authentication, voice authorization.