Real-Valued Sparse Bayesian Learning for Off-Grid DOA Estimation

摘要

Off-grid sparse Bayesian learning (SBL) direction-of-arrival (DOA) estimation methods exhibit many advantages, but they suffer from a high computational complexity. To reduce the computational complexity and improve the accuracy, we utilize a unitary matrix to transform complex manifold matrices into real ones and then use singular value decomposition (SVD) technique to reduce the dimension of matrices. Moreover, we consider the sampling grids as the adjustable parameters and adopt an expectation-maximization (EM) algorithm to reduce the modeling error iteratively. Since the conventional root refinement method is no longer suitable for the real-valued case, we utilize a fixed stepsize to update the locations of grid points. The simulation results demonstrate that our method can significantly reduce the computational complexity and improve the DOA estimation performance.