Influence of procedures for processing the voice signal of authentication system on the quality of formant data

Abstract

The current scientific task of improving the efficiency of voice authentication systems, which are widely used in modern access systems, is considered. It is known that it is possible to reduce errors of the first and second kinds in voice authentication systems by improving the procedures for the digital processing of the voice signal being analyzed, by better extraction of user features, or by improving procedures for decision-making on user admission. An important place in all voice signal processing procedures is occupied by formant data (spectral power levels, formant frequencies, spectral envelopes, and the width of the formant frequency spectrum). Based on the first two formants, speech recognition and synthesis are solved, and the next two formants enable user authentication. The purpose of this work is to outline ways to improve the quality of the formation of formant data in relation to the tasks of digital processing of speech signals. The object of the study is the process of obtaining formant data using amplitude-frequency and phase information, as well as the results of calculating the autocorrelation function of the analyzed signal. The subject of the study is methods and procedures for extracting formant data in the context of experimental research. The scientific novelty of the obtained research results lies in the fact that, for the first time, a comparative analysis of formant data obtained from different source information, namely amplitude-frequency and phase information, as well as the results of calculating the autocorrelation function of the analyzed signal, has been performed. The reliability of the research results is justified by the proper use of the known mathematical apparatus and the coincidence of formant data estimates as a result of processing the experimental user signal. The practical significance lies in the fact that the obtained results enable improvements in the quality and efficiency of voice data processing for speech recognition and synthesis, user authentication in voice systems, and several other applied tasks related to speech production.