Estimation of the number of signals based on a sequence of hypothesis test and random matrix theory

Estimating the number of sources impinging on an array of sensors is a well-known and a widely-studied issue. Source enumeration is typically a first step in blind source separation, detection of arrival and source localization tasks. The widespread approach for solving this issue is to use an information theoretic criterion like the minimum description length (MDL) introduced by Schwartz and Rissanen, or the Akaike information criterion (AIC). In this paper, we focus on a non-parametric approach where the behavior of eigenvalues of the sample covariance matrix is exploited. We present an estimator based on a sequence of hypothesis tests and recent results from random matrix theory (RMT). A series of simulations show its superiority over the classical estimators based on information theoretic criteria.