A Bayesian perspective on sparse regularization for STAP post-processing

Abstract

Abstract-Traditional Space Time Adaptive Processing (STAP) formulations cast the problem as a detection task which results in an optimal decision statistic for a single target in colored Gaussian noise. In the present work, inspired by recent theoretical and algorithmic advances in the field known as compressed sensing, we impose a Laplacian prior on the targets themselves which encourages sparsity in the resulting reconstruction of the angle/Doppler plane. By casting the problem in a Bayesian framework, it becomes readily apparent that sparse regularization can be applied as a post-processing step after the use of a traditional STAP algorithm for clutter estimation. Simulation results demonstrate that this approach allows closely spaced targets to be more easily distinguished.

Noise driven compressed sensing method for space time signal processing

“…In recent work related with compressed sensing based STAP [4,5,6,7], the coherence of the sensing matrix is not low due to the high resolution of the DOA-Doppler plane, which does not guarantee a good reconstruction of the sparse vector with large probability. Consequently, the direct estimation of the target amplitude may be unreliable using sparse representation when locating a moving target from the surrounding strong clutter.…”

“…In recent years, a number of compressed sensing based methods are proposed to detect unknown moving targets in strong clutter situation directly on the space-time data, which reduces the measurement data efficiently [4,5,6,7]. In [4], the entire radar scene, DOA-Doppler plane, is reconstructed using a compressed sensing based approach, and an attempt is then made to identify and zero out the clutter component.…”

“…However, it is based on the assumption of known clutter ridge location. The work in [6] is a combination of the traditional STAP algorithm and compressed sensing. The STAP is used to suppress the clutter, and the compressed sensing is then utilized to reconstruct the entire radar scene which only contains the targets.…”

Space-Time Adaptive Processing Based on Weighted Regularized Sparse Recovery

“…Recently, motivated by sparse representaton/sparse recovery (SR) techniques used in radar [5][6][7][8], several authors have considered SR ideas for moving target indication (MTI) and STAP problems, such as sparse-recovery-based STAP type (SR-STAP) algorithms in [9][10][11][12][13][14][15], L1-regularized STAP filters in [16,17], etc.. The basic idea of SR-STAP type algorithms is to regularize a linear inverse problem by including prior knowledge that the clutter spectrum is sparse in the angle-Doppler plane.…”

“…Under the assumption of the known clutter ridge in angle-Doppler plane, the authors in [10] imposed a sparse regularization based on l 1 -norm penalty to estimate the clutter covariance by excluding the clutter ridge. In [11], the authors presented a post-processing step after clutter whitening using a standard STAP technique by applying sparse regularization. To reduce the need for secondary data or for accurate prior knowledge of the clutter statistics, an iterative adaptive approach (IAA) is presented to compute the clutter covariance matrix in [12].…”