Частотно-часовий аналіз сигналів на основі функцій поведінки і арифметичних рядів: дослідження працездатності методу

Abstract

The article presents the results of studies on the health of the new method of time-frequency analysis of discrete signals, which are represented by time series. For estimation, a model signal is considered, formed as a sum of sinusoidal functions with known frequencies. As a result of using the method, spectrum estimates are obtained. The accuracy of the obtained spectrum estimates is estimated, as well as the accuracy of signal reconstruction based on this spectrum for a model signal.According to the results of the study, we can conclude that the proposed method allows us to estimate the signal spectrum with a fairly high accuracy even on the basis of a limited amount of data. In many ways, the accuracy of the approach will depend on many parameters of the signal processing algorithm. In particular, from the conditions of signal sampling due to the operation of analog-to-digital converters of real devices, the parameters of the algorithm for constructing the behavior function and identification of the metasystem, the window size for generating the balance equation, the gain of the fuzzy filter, and other parameters. These parameters are the settings of the time-frequency signal analysis algorithm based on the proposed method.Studies have shown that in the case of considering signals having high-frequency components, it is advisable to lower the cutoff threshold when identifying the met system. The need to identify an unsteady signal spectrum requires an increase in the fuzzy filter gain. The analysis of the algorithm also showed other possibilities of its adjustment, which make it possible to increase the efficiency of the proposed method of time-frequency analysis of signals.