一种改进的分段平稳随机过程的参数估计方法

AN ADVANCED METHOD TO ESTIMATE PARAMETERS OF PIECEWISE STATIONARY STOCHASTIC PROCESS

将非平稳随机信号划分为分段平稳随机信号进行处理,为非平稳随机信号的研究提供的一种新的分析方法。为最优地将非平稳随机信号划分为分段平稳随机信号,Djuric等人用 Bayes方法,通过递推多维条件分布概率来估计最优划分参数值,但计算相当复杂。本文在研究 AR模型本身的一些特性的基础上,通过直接递推多维联合分布概率来估计最优划分参数,大大地减少了计算量。

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 process. Some authors estimated the optimum parameters through calculating recursively the multivariate conditional likelihood function, which made the computation very complex. Basing on some natural characteristics of Aft mode, a new recursive method is provided, which can improve the computation efficiently, to estimate the optimum parameters.