Off-grid DOA Estimation using Temporal Block Sparse Bayesian Inference

By considering the off-grid distance in the sparse reconstruction model, off-grid direction-of-arrival (DOA) estimation can achieve better performance. Most existing off-grid algorithms consider that the snapshots of each source are independent with each other. This contradicts with the real-world scenario, where sources often have temporal structures. To address this issue, we present a new off-grid DOA estimation method, which brings the temporal structures into the off-grid model and a temporal block sparse Bayesian inference is derived. In comparison with the off-grid block sparse Bayesian inference method, the proposed approach achieves higher estimation accuracy for off-grid source directions in low SNR situations. Numerical simulations demonstrate the preferable performance of our method.