在用辅助变量法对SISO线性模型进行辨识时,如果将其数据向量和辅助向量重新排列并扩维,利用由此而出现的“移位性质”,则可构成一种新的数据乘积矩矩阵,其逆矩阵称作信息压缩阵。本文提出一种新的UDV~r分解方法,并利用它对这个信息压缩阵进行递推分解,构成了一种能同时进行模型参数和阶次辨识的递推算法。使用这种新的辨识算法,可以减小整个辨识过程的计算量,改善数值计算品质,提高辨识精度。 When using the auxiliary variable method to identify the SISO linear model, if the data vectors and auxiliary vectors are rearranged and expanded, the resulting “shift property” can be used to form a new data product matrix Its inverse matrix is ​​called information compression matrix. In this paper, we propose a new UDV-r decomposition method and use it to recursively decompose the information compression matrix to form a recursive algorithm that can simultaneously model parameters and order identification. Using this new identification algorithm, the computational cost of the entire identification process can be reduced, the numerical calculation quality can be improved, and the identification accuracy can be improved.