About one algorithm of determining the frequency intervals information

Abstract

Considers the problem of analysis of speech signals, with the aim of identifying their characteristics. The needs to solve these problems arise in the field of speech recognition and segmentation, its compression and removal of noise. Analysis of the literature shows that the most commonly used frequency representation in solving such problems. The most effective is the use of sub-band frequency analysis, the essence of which lies in the correlation properties of the signal segments with some splitting the frequency axis into intervals of finite width. In this article the results of research method for determining the frequency domain information in the speech signals obtained on the basis of sub-band frequency analysis. The dependence of the probability of an erroneous definition of frequency intervals information on the energy level of noise overlay. Results of the study algorithm for determining the frequency intervals of information indicate the feasibility of using an adaptive threshold to determine the frequency intervals of information for the analysis of speech signals.