A new algorithm for source enumeration in large dimensional regime
Shoucheng Yuan, Bin Zhang, Junjie Wang · 2023
This article proposes a method for estimating the number of signals with high-dimensional low sample size. The inspiration for the method comes from testing whether the covariance matrix is spherical when the sample size is less than the dimension. The test statistic consists of the first four moments of sample eigenvalues and relaxes the assumption of Gaussian distribution. Based on the generalized BIC, the expression for determining the number of source signals is given. Simulation results demonstrate that the proposed method has a high probability of detection in both the Gaussian and the non-Gaussian noises and performs better than some existing methods.