Performance analysis of estimation methods for incoherently distributed source number

We consider the problem of estimating the number of sources by means of a passive sensor array for distributed sources. This scenario occurs in mobile communications when local scatterers exist in the vicinity of a transmitter. Since the distributed source brings about a full rank data model, the performance of eigenstructure-based detection criteria deteriorates, especially at high signal to noise ratio (SNR). More specifically, the detection criteria overestimate the number of sources at high SNR. We propose a robust detection criterion through a pseudo-procedure which allows us to restore the rank of a signal subspace for a uniform circular array. This criterion is based on a minimum description length (MDL) representing the asymptotically consistent characteristic. We show the improved performance and the robustness of the proposed method for distributed sources.