基于特征值聚类的MUSIC算法

MUSIC algorithm based on eigenvalue clustering

传统MUSIC算法需要预先知道目标信号源的个数,进一步确定信号子空间和噪声子空间的维数,最后进行谱峰搜索。在实际工程中,无法预知待测目标的个数,针对这一问题,提出了一种基于密度聚类算法的改进型MUSIC算法。该算法将协方差矩阵的特征值进行聚类,通过DBSCAN聚类算法可求出目标信号源的个数,再进一步估计出目标的方位。仿真结果表明:提出的改进算法在信号源个数未知的情况下能够准确估计出信号源的个数和方位,较传统的MUSIC算法有更大实用性。

The traditional MUSIC algorithm needs to know the number of target signal sources in advance,and further determine the dimensions of signal subspace and noise subspace,and finally search for spectral peaks.In engineering,it is impossible to predict the number of target signal sources to be measured.To solve the above-mentioned problem,an improved MUSIC algorithm without estimating the number of target signal sources is proposed.In the present algorithm,all eigenvectors of covariance matrix are regarded as noise subspace for spectral estimation,but the existence of signal subspace will make the result unreliable.In order to make the estimation result more accurate,a new weighting method for the spectral estimation results of noise subspace and signal subspace is proposed.The simulation results show that the improved algorithm can accurately estimate the number and direction of signal sources when the number of signal sources is unknown,and has greater practicability than the traditional MUSIC algorithm.In addition,the improved algorithm has better robustness.