摘要
在文[6]研究的基础上.把文[6]中提出的偏差补偿最小二乘法(BELS)推广到多变量线性系统的辨识中.分析表明,只要有色噪声与系统的输入信号统计不相关,我们总可以通过对输入信号预滤波的方法将已知零点嵌入到被辨识系统中,然后利用这些零点提供的信息消除噪声引起的辨识偏差.这种方法的特点是其辨识过程不依赖于有色噪声的模型.
n this paper, the Bias-eliminating Least-squares method (BELS) proposed in[6] is extended to the identification of multivariable systems. Designed filters are connected to the system at the input terminals.so that the augmented system has some known zeros which can be used for eliminating the noise-induced bias from the ordinary least-squares estimators. It has been shown that the presented method can provide consistent parameter estimation even if there is not any priori knowledge about the noise. Results of simulation examples verify the theoretical analyses.
出处
《控制与决策》
EI
CSCD
北大核心
1995年第3期265-269,共5页
Control and Decision
基金
国家自然科学基金
关键词
多变量系统
参数估计
偏差校正法
最小二乘法
parameter estimation, least-square method, linear multivariable system