Abstract:Reviews on the Evolutionary Power Spectrum Density (EPSD) estimation of stochastic process via wavelets are presented in the paper. Based on a newly proposed Local Stationary Wavelet (LSW) model of non-stationary stochastic process, an approach of estimating the time-varying PSD is developed. The newly proposed approach can be explained in a unified perspective with the classic one. Both the approaches reduce to a same form when the general harmonic wavelet is applied. The auto/cross-EPSD of a multi-variant stochastic process is employed as a numerical example to demonstrate the efficiency of the approach. A real world situation, including both the near-field and far-field ground motions of the Wenchuan, China (05/12/2008), are applied as an example to show the applicability of the time-frequency analysis of the approach.