时变AR模型阶数确定与系数估计的方法  

Order determination and parameter estimation of the time-varying AR model

作　　者：高伟伟[1] 申丽然 

机构地区：[1]哈尔滨工程大学信息与通信工程学院,黑龙江哈尔滨150001

出　　处：《应用科技》2010年第11期30-34,共5页Applied Science and Technology

基　　金：国家自然科学基金资助项目(60803087)

摘　　要：研究了用时变自回归(TVAR)模型对非平稳信号建模的方法.对该模型进行详细分析,探讨了参数模型辨识存在的2大问题:模型阶数的确定和基函数的选择.基于现定阶准则只适用于短时平稳信号的分析,所以利用具有时变特性的信息理论准则(information theoretic criteriaI,TC)来确定模型的阶数.通过引入基函数,利用最小二乘算法对模型系数进行估计,从而将非平稳信号的时变模型转化为线性时不变模型,并比较了几种基函数的拟合性能.证明了由于墨西哥草帽小波基函数具有良好的时频特性并且在使用时无需预知信号的先验信息,从而优于其他传统的基函数.The application of time-varying autoregressive （TVAR） modeling approach to non-stationary signals was studied. The model was analyzed and two major problems existed in parametric model identification were discussed： the determination of model order and the selection of primary function. The commonly used order determination criterion at present only applies to analysis of short time stationary signal, so the information theoretical criteria （ITC） was used to determine the order of the model The parameters were estimated by using the recursive least square algorithm of the primary function, and thereby the time-varying model of non-stationary signals can be translated into a linear timeinvariable problem. Several kinds of primary functions were analyzed for their fitting functions, which proved that the Mexican hat function has good time-frequency characteristics and it does not need prior information when using the function, so it is better than other traditional primary functions.

关 键 词：TVAR 时变参数模型 非平稳信号 时变参数估计 

分 类 号：TN911.6[电子电信—通信与信息系统]