Time and lag window selection in Wigner-Ville distribution

The paper provides the conditions on the selection of the lap and the time windows in Wigner-Ville distribution when it is considered for time-varying power spectrum estimation. It is shown that the general class of non-stationary processes requires the inseparability of the two windows. The separation is adequate for a certain class of processes in which the autocorrelation is considered invariant along time intervals whose length is shorter for smaller lags. The paper presents a discussion which relates the time-averaging estimation of the autocorrelation function with the nature of its slow time-variation.