Результати попередньої обробки голосового сигналу в системах автентифікації

Abstract

The report examines the current scientific problem of preprocessing a voice signal in authentication systems, which is currently implemented in the field of amplitude-frequency characteristics of the processed data. To increase the efficiency of voice signal preprocessing procedures, it is proposed to use its phase data. The phase data of a voice signal in the time domain is known to be in the form of sawtooth signals of unknown duration. It has been established that during the phase change period there are one or two harmonics. The selection of signals with one harmonic can be performed using linear approximation and the chi-square criterion. Further research will be focused on identifying the degree of the polynomial for phase signals with two harmonics.