Online Sequential Compressed Sensing With Multiple Information for Through-the-Wall Radar Imaging

Abstract

We propose a novel strategy for applying compressed sensing (CS) to stepped frequency continuous wave synthetic aperture radars. The measurements are performed by adhering to a sequential measurement strategy. The sensor autonomously adapts, on the fly, the number of samples needed to reconstruct the reflectivity function and guarantees an imposed reconstruction quality. The measurements obtained at the previous scanning positions are added as side information into the reconstruction of the reflectivity function sensed by the radar at the current radar position. The algorithm attributes autonomous weights to each of these side information depending on the similarity with the signal to reconstruct. This approach is tested and evaluated on a series of simulated and real through-the-wall imaging radar measurements for detecting static human targets hidden behind a wall. The experiments first prove that the frequency sampling rate can be decreased far below the bound obtained by the common CS approach and, second, that the algorithm allows determining an accurate upper bound for the reconstruction error, and thus to autonomously decide online on the number of samples.