On spectral analysis with nonuniform frequency resolution of nonstationary stochastic processes

Abstract

Spectral analysis with nonuniform frequency resolution of nonstationary stochastic processes is addressed. The frequency-warping operation aimed at increasing the frequency resolution is shown to modify the nonstationarity kind of the analyzed process. Specifically, in several cases of interest, the frequency-warped process is shown to belong to the recently introduced class of the spectrally correlated processes. Spectral correlation density estimation is performed by frequency smoothing the periodogram along curves in the bifrequency plane instead of lines with unit slopes as in the case of wide-sense stationary and almost-cyclostationary processes. Application to cyclic spectral analysis of the GPS-L1 signal is considered.