A Sparse Learning Based Detector with Enhanced Mismatched Signals Rejection Capabilities

This paper devises a tunable detection architecture to deal with mismatched signals embedded in Gaussian interference with unknown covariance matrix based on a sparse recovery technique. Specifically, a sparse learning method is exploited to estimate the amplitude and angle of arrival of the possible targets, which are then employed to design detectors relying on the two-stage detection paradigm. Remarkably, the new decision scheme exhibits a bounded-constant false alarm rate property. The performance assessment, carried out through Monte Carlo simulations, shows that the new detectors can outperform classic counterparts in terms of rejecting mismatched signals, while retaining reasonable detection performance for matched signals.