基于改进的K-均值聚类信源数目估计算法

针对盖尔圆盘法在低信噪比和少快拍数背景下进行信源数目估计时会出现错误判断的问题,提出了色噪声背景下改进的K-均值聚类信源数目估计方法.求出数据相关矩阵后按照信号和噪声的类均值对信号特征值和噪声特征值进行K-均值聚类,然后不断更新类均值直到信号特征值收敛.计算机仿真实验证明了该方法在低信噪比和少快拍数时的估计成功概率高于盖尔圆盘法.

Estimation Method for the Number of Sources Based on Modified K-means Clustering

In view of the problems of inefficient in low SNR and less snapshots when using signal source number estimation method of Gerschgorin disks(GDE),an improved K-means method based on clustering algorithm under colored noise background was proposed to estimate number of signal sources.After getting the correlation matrix,we constructed a distinguish criteria using K-means clustering in reason according to the difference between the signal and noise eigenvalues,then signal and noise eigenvalues were classified ...