一类Markov跳变系统的综合时频分析模型切换检测

Integrated Time-frequency Analysis to Detect Model Switch of a Class of Markov Switch Systems

在Markov跳变系统的估计问题中,目前的多模型估计器采用模型概率检测模型切换;这是一种基于信号幅值的检测方法;事实上,模型切换表征了系统的暂态过程,因而系统某些特征的频谱特性在模型切换前后存在差异;因此提出一类Markov跳变系统的综合时频分析模型切换检测方案(ITFA);同时给出统计意义下的模型检测评价指标;仿真结果表明:与标准的交互式多模型算法相比,ITFA方法具有检测概率高、延迟时间小的优点;同时,ITFA对于先验性的滑窗长度、检测阈值具有较好的鲁棒性。

In the estimation of Markov switch systems, the well known multiple model estimators （MME） detect model switch based on model probability. It belongs to the amplitude based detection method. In fact, model switch represents the system transient process. Hence there exists a certain signal with significant difference in frequency domain before and after model switch. Based on the idea of integrated time-frequency analysis （ITFA）, a new scheme to detect model switch is proposed for a class of Markov switch systems. In addition, several performance indexes about detection are given, In computer simulation of maneuvering target tracking, the ITFA is shown higher detection probability and smaller detection delay, compared with the standard Interacting Multiple Model （IMM）. Besides this, the ITFA also has strong robustness to a priori parameters, such as moving window size and detection threshold.