一种修正极小方差谱估计算法

A Modified Minimun Variance Spectral Estimation Algorithm

通过分析极小方差 (MV)谱估计算法与多信号分类 (MUSIC)谱估计算法之间的关系 ,揭示出在对样本的自相关矩阵进行特征分解后 ,对各特征矢量权系数的不同选择决定了二者性能上的差异。在此基础上 ,构造了一种新的权系数 ,导出了修正极小方差 (MMV)谱估计算法。仿真结果表明 ,MMV算法的分解率高于MV ,方差小于修正协方差自回归 (AR)模型谱估计算法 ,并能正确反映正弦信号间功率的相对大小。

By exploiting the relation between minimum variance(MV) and multiple signal classification (MUSIC)spectral estimation algorithms,it is shown that different choices of cofficients for the eigenvectors of autocorrclation matrix lead to a difference in their performances This paper prosases a new kind of cofficients and accordingly,derives a modified minimum variance(MMV) spcctral estimation algorithm.The simulation results have revealed that the MMV algorithm has a better resolution than that of MV and is more robust to noise than AR. Moreover,it can hold the magnitade of the sine signals'power in an accurate manner.