Approximate multitarget matched filter for MIMO radar detection via Orthogonal Matching Pursuit

Abstract

In this article, we consider the problem of detecting multiple targets in MIMO radar. MIMO ambiguity functions generally present strong range/angle coupling or high sidelobe levels so that weak targets will often be buried in the sidelobes of stronger targets. We propose to solve this problem by iteratively building an approximation of the multitarget matched filter through an Orthogonal Matching Pursuit (OMP) procedure that permits to clean the received signal from the strong target sidelobes and thus to detect the low-SNR targets. However, whereas classic OMP exploits a finite discretization of the target parameter space, this procedure must be adapted here to deal with real radar targets that can be located anywhere in the range/angle/doppler cell. We propose here to solve this problem by jointly estimating at each iteration the target states in the maximum likelihood sense via a gradient descent algorithm. We also show that the stopping criteria for this iterative procedure can be set to satisfy a given false alarm probability. Simulations show that the proposed method permits to solve the multitarget detection problem in the MIMO framework and retrieve low-SNR targets buried in the high sidelobes of strong targets produced by the MIMO matched filter.