一种适用于小样本的迭代多重信号分类算法

当样本数不足时，由采样协方差矩阵特征分解得到的噪声子空间偏离其真实值，使得多重信号分类(MUSIC)算法目标角度(DOA)估计性能下降。为了解决这个问题，该文提出了一种迭代算法通过校正信号子空间来提高MUSIC算法性能。该方法首先利用采样协方差矩阵特征分解得到的噪声子空间粗略估计目标角度；其次基于信源的稀疏性和导向矢量的低秩特性，由上一步得到的目标角度以及其邻域角度对应的导向矢量构造一个新的信号子空间；最后通过解一个优化问题来校正信号子空间。仿真结果表明，该算法有效地提高了子空间估计精度。基于新的信号子空间实现MUSIC DOA估计可以使得性能得到改善，且在低样本数下改善尤为明显。

Iterative Multiple Signal Classification Algorithm with Small Sample Size

For cases with small samples, the estimated noise subspace obtained from sample covariance matrix deviates from the true one, which results in MUltiple SIgnal Classification (MUSIC) Direction-Of-Arrival (DOA) estimation performance breakdown. To deal with this problem, an iterative algorithm is proposed to improve the MUSIC performance by modifying the signal subspace in this paper. Firstly, the DOAs are roughly estimated based on the noise subspace obtained from sample covariance matrix. Then, considering the sparsity of signals and the low-rank property of steering matrix, a new signal subspace is got from the steering matrix consisting of estimated DOAs and their adjacent angles. Finally, the signal subspace is modified by solving an optimization problem. Simulation results demonstrate the proposed algorithm can improve the subspace estimation accuracy and furtherly improve the MUSIC DOA estimation performance, especially in small sample cases.