An estimator of the number of sources based on a sequence of hypothesis test
Manlin Xiao, Janqi Lu, Ping Wei · 2011 International Conference on Computational Problem-Solving (ICCP) · 2011
Estimation of the number of sources embedded in noise is a fundamental problem in statistical signal and array processing. This paper focuses on a non-parametric tool to estimate the number of sources without any information about the signature matrix. We exploit the behavior of eigenvalues of the sample covariance matrix and propose a new estimator based on a sequence of hypothesis test. A series of simulations show its superiority compared to the classical estimators based on information theoretic criteria.