Spectral smoothing and recursion based on the nonstationarity of the autocorrelation function

Abstract

Averaging and recursion in short-time spectral analysis in view of the nature of the nonstationarity of the random process is addressed. Using the Wigner spectrum estimator as a general model, this correspondence relates the selection of the time and lag windows to the stationary intervals of the autocorrelation at different lags. The lag-dependent time averaging for two types of commonly assumed nonstationarities is presented and used to devise the recursive estimates of their time-varying spectrum. >