UDC 621.396; 621.391.82; 621.37:51-74
PARAMETRIC SPECTRAL ANALYSIS OF PIECEWISE-STATIONARY RADIOENGINEERING SIGNALS WITH VARIOUS CORRELATION PROPERTIES
V. G. Andrejev, Dr. Sc. (Tech.), full professor, the department of Radioengineering systems, RSREU, Ryazan, Russia;
orcid.org/0000-0003-3059-3532; e-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.
V. A. Tran, post graduate student, RSREU, Ryazan, Russia;
orcid.org/0000-0002-6743-0131, e-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.
We proposed and investigated a modified method of spectral analysis of piecewise-stationary processes to take into account the influence of additive noise changing power on correlation matrices. The aim of the work is to increase the computational efficiency of analysis algorithms and the accuracy of spectral estimation of radioengineering signals on the background of piecewise-stationary noises. In the process of estimating the optimal value of weight coefficient β, which determines the shares β and (1−β) of processes with different statistical properties the method proposed makes it possible to reduce the influence of non-stationary noise and improve the accuracy of spectral estimates by correcting autocorrelation coefficients of piecewise stationary random processes. The qualitative indicators of a proposed modified spectral analysis method are compared with a conventional parametric autoregressive method. Experimental studies have shown that the application of the proposed approach for spectral estimation, when compared to known autoregressive methods, allows reducing the discrepancy between control and estimated spectra by 3.1...5.5 times. When conducting a comparative analysis with a conventional autoregressive model, the decrease in the order of p can reach 2...3 times while maintaining the same spectral estimation accuracy. It is confirmed that for spectrum analysis of narrowband radioengineering signals under consideration, relative deviations ΔF of dominant frequency estimate are significantly (up to 4 times) reduced by using a proposed modified method in comparison with an autoregressive method. Winnings in the accuracy of spectral estimation are achieved by taking into account changes in autocorrelation properties of the analyzed signal due to the effect of piecewise stationary noise on it, the power of which being changed during observation
Key words: wise-stationary noise, non-stationary noise, weight coefficient, adaptive algorithm, autoregressive model, change-point, spectral estimation, power spectral density