摘要
提出了一种新的递归全局最小二乘快速算法，其可用于递归计算自适应（ＦＩＲ）滤波问题的全局最小二乘（ＴＬＳ）解．在这个算法中，以增广数据矢量为优化搜索方向，其快速计算归结为新定义的增益矢量的快速计算．利用数据矢量的位移结构找到了计算增益矢量的快速算法，它的运算量比Ｋａｌｍａｎ增益矢量的快速算法的运算量少，且不存在数值计算不稳定性问题．在追踪与增广协方差矩阵的最小特征值相关联的特征矢量过程中，这个算法每步须Ｏ（Ｎ）乘法运算，比Ｄａｖｉｌａ的ＲＴＬＳ算法的运算量少．
A fast recursive total least squares algorithm is developed to allow recursive computation of total least squares solution for adaptive finite impulse response (FIR) filters. The augmented data vector is used as search vector for the iterative algorithm. Using the shift structure of the augmented data vector, the fast algorithm for the gain vector is given. Unlike fast algorithm of Kalman gain vector, less operations per time step is necessary, while being stable numerically. The operations for extracting the smallest eigenvalue and the corresponding eigenvectors of the augmented data correlation matrix is O(N) and less than the ones of Davila′s RTLS algorithm per time step. The performance of the algorithms is evaluated via simulations.
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
1999年第4期44-47,51,共5页
Journal of Xi'an Jiaotong University
基金
国家自然科学基金
