Wold-Cramer evolutionary spectral estimators

A nonstationary signal can be represented as the output of a linear time-varying system, characterized by Zadeh's system function, with white noise as input. The Wold-Cramer evolutionary spectrum is defined as the magnitude squared of this system function. Estimators are obtained by expressing the signal as a random time-varying amplitude modulating a complex exponential of a certain analysis frequency, by representing this amplitude function by an expansion of orthonormal functions, by computing a minimum-mean-square-error estimate of the amplitude function, and by repeating the estimation over all frequencies. Estimators analogous to the stationary periodogram and Capon's minimum-variance estimators are obtained. They compare favorably with traditional estimators, as well as with recent techniques.<
>