基于矩陣正交分解的自适应算法

自适应算法是各种自适应系统的核心.计算稳定性是衡量算法优良程度的重要指标之一.本文讨了论两类最小二乘算法的计算稳定性.其一是基于正则方程的递推求解方法,这是目前各种自适应算法的基础;另一是正交分解算法.后者的计算稳定性要比前者高.研究了一种基本的正交分解自适应算法,并指出,如果初始条件恰当,它将得到和批处理同样的结果.将这种方法用于空间谱估计,初步结果表明,用它能实现高分辨率精确测向.

ADAPTIVE ALGORITHM BASED ON MATRIX ORTHOGONAL DECOMPOSITION

The adaptive algorithm is the kernel of adaptive, systems. The comput ational stability is one of the most important characteristics of algorithms. In this paper, we discuss the computational stability of two kinds of algorithms for solving Least-Squares problem, the normal equation method and the orthogonal decomposition methdd, and show that we can get better performance in Jater case in this aspect. An adaptive algorithm based on matrixo rthogonal decomposition has been developed for AR parameter estimation. It has been shown that the same results would be obtained by using this algorithm as those obtained by underbatch paocessing, provided that the appropriate initial conditrons are met. Preliminary simulation results show that the high resolution Direction-Finding can be estabiished with the application of this method to spatial spectral estimation.