A source number estimation method based on total least squares fitting

摘要

The color noise will lead to performance degradation of the most source number estimation methods. Focus on this problem, a novel source number detection method used in colored noise is proposed in this paper. First, the method is based on eigen-decomposition of array covariance matrix and then sorts the eigenvalues in descending order. Though a straight line of the total least-squares (TLS) fitting with the eigenvalues after eigen-decomposition, and using the characteristics of the minimum distance sum of square between the eigenvalues and its fitting straight line, a novel estimation function is constructed. By finding the minimum value of the target function, the source number estimation value could be acquired. The method can get a higher detection probability in colored noise, the simulations and the experiment data proved it correctly.