正交最小二乘算法及其应用

介绍了正交最小二乘法的算法原理，同经典的最小二乘法相比，其数值稳定性好、计算量小，能节省大量存贮空间。应用了Ｈｏｕｓｅｈｏｌｄｅｒ变换、Ｇｉｖｅｎｓ变换和它的逆变换３种算法。根据残差矢量最小准则，将该法用于ＡＲＭＡ模型结构辨识和时变ＡＲ模型参数估计，取得了满意的仿真结果。

Orthogonal least-squares estimation and its applications

The algorithm principle of the orthogonal least-squares estimation is dealed with. Comparing with the classical least-squares estimation, this algorithm is characterized by its better numerical stability, less computation and saving more store space. The Householder transformation is applied in the algorithm 1. The Givens transformation and its inverse transformation are applied in the algorithm 2 and algorithm 3, respectively. Based on the minimum of the residual error vector, this approach is applied to the structure identification and parameter estimation of ARMA models, as well as to the parameter estimation of time varying AR models. Some successful results have been obtained in simulation.