MUSIC算法是一种基于矩阵特征空间分解的方法。从几何角度讲，信号处理的观测空间可以分解为信号子空间和噪声子空间，显然这两个空间是正交的。信号子空间由阵列接收到的数据协方差矩阵中与信号对应的特征向量组成，噪声子空间则由协方差矩阵中所有最小特征值（噪声方差）对应的特征向量组成。(MUSIC algorithm is a kind of feature space based on the matrix decomposition method.From geometric point of view, the signal processing of the observation space can be decomposed into signal subspace and noise subspace, obviously the two space are orthogonal.Signal subspace by array covariance matrix of the received data and signal of the corresponding eigenvectors, the noise subspace by the covariance matrix of all minimum eigenvalue (noise variance) of the corresponding eigenvectors.)