PERTURBED COMPRESSED SENSING BASED SINGLE SNAPSHOT DOA ESTIMATION

Traditionally, direction of arrival (DOA) estimation techniques have been based on spectral estimation methods utilizing signal and noise subspaces [1]. Such techniques perform well when sensor measurements are available at multiple snapshots. Recently, compressed sensing (CS) based DOA estimation techniques have been introduced, which improve source localization in the single snapshot case by modeling the angle search as a sparse recovery problem. In this domain, various on-grid and off-grid methods have been proposed in the existing literature [2] [3] [4]. The on-grid methods rely on a fixed basis and solve traditional CS based sparse recovery problems while the latter has modifications based on first-order Taylor approximation of the array manifold matrix. In this paper, we present an off-grid CS based formulation, where we employ an alternating minimization strategy for fine-grid search of source directions based on coordinate descent. We show that our technique outperforms the first-order approximation techniques whose performance is limited by the signal-norm dependent Taylor error.