Abstract：A new way to analysis nonstationary stochastic process is to divide it into piece-wise stationary stochastic process. Djuric(1992) used Bayes method to estimate the parameters, which can optimally divide the nonstationary stochastic process into stationary stochastic pro-cess. Some authors estimated the optimum parameters through calculating recursively the multivariate conditional likelihood function, which made the computation very complex. Bas-ing on some natural characteristics of AR. mode, a new recursive method is provided, which can improve the computation efficiently, to estimate the optimum parameters.
2003, Vol. 25 
