基于线性噪声特征值的信源数估计

Sources Number Estimation Based on Linear Noise Eigenvalues

为了提高信源数目估计的准确性,本文通过对采样协方差矩阵特征值的分析,得出噪声子空间对应的特征值近似线性而信号子空间对应的特征值不在该直线上的性质,根据此性质提出一种简单有效的基于最小二乘直线拟合的信源数估计算法,可以在低信噪比的情况下准确估计信号源数目。

In order to improve source number estimation accuracy, a characteristic that eigenvalues of noise subspace are approximately linear but the eigenvalues of the signal subspaee are not on the line is obtained through analyzing eigenvalues of sampling covariance matrix. According to this feature, a simple and efficient source number estimation algorithm based on least square line fitting was proposed, simulation results show that it can estimate the number of signal source at low signal-noise ratio （SNR） accurately.