基于矢量奇异值分解的DOA估计方法及其改进

摘要

对相干信号的波达方向（DOA）估计是空间超分辨谱估计的热点。在均匀线性阵列模型下，特征矢量奇异值分解法（ESVD）能够很好的对相干信号进行DOA估计，但是当相干信号和非相关信号同时存在时，ESVD并不能对全部信号进行DOA的估计。本文通过对ESVD算法的理论分析后，选取经过加权处理的特征向量来构造新矩阵，再利用奇异值分解得到信号的噪声和信号子空间，从而进行DOA估计。理论分析和计算机仿真表明该改进算法(MESVD)解决了ESVD算法在相干信号和不相关信号同时存在不能正确进行DOA估计的问题，估计精度与空间平滑算法（FBSS）相当。%Abstrct:The direction of arrival (DOA) estimation of coherent signals is a hotspot issue of High-resolution spatial spectrum estimation.Under the Uniform Liner Array model,the Extended Signal Value Decomposition(ESVD) algorithm can estimate the DOA of coherent signals exactly.However, when the coherent and non-related signals existing at the same time,ESVD fails to estimate al DOA of the signals. Through the theoretical analysis of the ESVD algorithm.a modified algorithm is proposed which select the weighted eignvector to construct a matrix for subspaces estimation. Theoretical analysis and computer simulation indicate that the improved algorithm (MESVD) solves the defect of the ESVD algorithm that it can not estimate al DOA of the signals when the coherent and non-related signals existing at the same time. The estimation accuracy of MESVD resembles the FBSS algorithm.