Sparse Bayesian learning for off-grid DOA estimation with nested arrays

The existing off-grid sparse Bayesian learning (SBL) DOA estimation method for nested arrays suffers from two major drawbacks: reduced array aperture and high modeling error. To solve these issues, a new data model formulation is first presented in this paper, in which we take the noise variance as a part of the unknown signal of interest, so as to learn its value by the Bayesian inference inherently. Then, we provide a novel grid refining procedure to eliminate the modeling error caused by off-grid gap, where we consider the locations of grid points as adjustable parameters and proceed to refine the grid point iteratively. Simulation results demonstrate that our method significantly improves the DOA estimation performance especially using a coarse grid.