Non-stationary stochastic signal was divided into piecewise stationary stochastic signal

and reflecting the sig-nal’s characteristics by autocorrelation function of the piecewise stationary stochastic signal.Generally

the autocorrela-tion function was the base of selecting signal base for signal representation.For expressing non-stationary stochastic sig-nal in a precise and effective way

based on the analysis of the natural characteristics of ARMA model and Markov proc-ess

a kind of multi-parameter estimation model of autocorrelation function for piecewise stationary stochastic process was proposed.The computational complexity was reduced

and the approximation effect was improved.Furthermore

the multi-parameter estimation model could also be adapted to the complex non-stationary stochastic signal

The computer simulation demonstrates that the approximation error was decreased significantly.