Specifics of Receiving and Processing Phase Information in Voice Authentication Systems

The issues of improving the reliability of storing various resources, access to which is carried out using telecommunication networks, are considered. In this case, the first barrier in ensuring access reliability is the user authentication system. Lately, access systems based on biometric features of a user have been used. Initially, static biometric features of a user (facial image, finger papillary pattern and iris) were preferable, which did not meet the expectations of developers and users due to the simplicity of their falsification. Recently, the preference has been given to the dynamic (behavioral) biometric features of a user, namely voice authentication systems became more widely used. As it is known, voice authentication systems have several advantages, such as: simplicity, convenience, compactness, low cost, and a number of others. In addition, the passphrase can be quickly changed and expanded during the authentication process. However, the quality indicators of all biometric access systems do not meet the increasing requirements. The object of the study is the process of digital processing of voice signal during user authentication in access systems.In the process of voice authentication, the analysis of the amplitude-frequency spectrum of recording materials is performed. At the same time, the main research focuses on the use of estimates of formants, cepstral coefficients, mel-frequency cepstral coefficients, linear prediction coefficients, etc. as a user’s template. On the basis of user’s established patterns, admission decisions are made using Gaussian Mixture Models, Support Vector Machines, Hidden Markov Models or artificial neural networks.In the report, it is proposed to change the paradigm of digital processing of user voice signals and supplement the analysis of the amplitude-frequency spectrum with studies of phase data, which are traditionally ignored during the authentication. According to the authors, the latter is caused by the lack of effective procedures for the formation of phase data, the requirement of additional computational resources, which were not always available to researchers, and some features using the signal phase.