摘要
就利用衰减因子通过指数加权对历史数据进行渐消记忆的时变参数估计问题,提出基于正交变换的递推最小二乘法,避免了传统方法因求解法方程而导致的病态加剧问题.同时降低了运算量,改善了收敛性,节省了存贮空间.此外,以本文递推最小二乘法为基础,还可构成相应的递推广义最小二乘.多级最小二乘、增广矩阵法、极大似然法等一系列时变参数渐消记忆估计方法.
A on-line least squares identrfication method is proposed for the parameter estimation of a timevariable system by introducing a filter factor to suppress exponentially previous measurements.The method is developed based on Houscholder transformation rather than normal equation of the over—de- termined equation so as to avoid deteriorating the ill-condition of the matrix.The fewer arithmetic operations, less storage requirements,and the betterconvergence are another three benefits of the method.In addition,the proposed on-line leastsquares method can easily be extended to develop a serious of corresponding methodsbased on Houscholder transformation, such as the on-line generalized least squares method,the on-line multiple stage least squares method,the on-line argmented matrix method,the on-line maximum likelihood method and so on.
出处
《信号处理》
CSCD
北大核心
1991年第4期220-228,共9页
Journal of Signal Processing
