改进的实对称阵特征值分解并行化算法

An Improved Parallel Algorithm for Eigenvalue Decomposition of Real Symmetry Matrix

实对称阵特征值分解在阵列信号处理中有十分重要的作用,例如波达方向估计技术中的MUSIC算法的主要计算量就是在对接收矢量的协方差矩阵进行特征值分解,因此采用并行化算法有重要意义。分析MUSIC算法的硬件实现特点后,提出了一种实对称矩阵特征值分解的并行化算法。采用该算法硬件实现时,并不增加计算复杂度,且计算机仿真结果表明在相同精度要求下,该算法的效率比常用算法明显提高。

Eigenvalue decomposition of real symmetry matrix plays an important role in array signal process.For instance,eigenvalues decomposition of the covariance matrix of received array accounts for the major computation of MUSIC algorithm.As a result,it is significant to apply parallel algorithm.Based on analysis of MUSIC hardware implementation,this paper proposes a new parallel algorithm for eigenvalue decomposition of real symmetry matrix.The application of this algorithm for hardware implementation gives no increase in computation complexity,and the simulation indicates that it is much more efficient than the common algorithms.