Root Sparse Bayesian Learning-Based 2-D Off-Grid DOA Estimation Algorithm for Massive MIMO Systems

Traditional sparse Bayesian learning (SBL)-based two-dimensional (2-D) direction of arrival (DOA) estimation algorithms exhibit limited accuracy in low signal-to-noise ratio (SNR) and small snapshot scenarios in massive multiple-input multiple-output (MIMO) systems. Therefore, we propose an improved DOA estimation algorithm based on off-grid root SBL and pseudo-noise resampling for massive MIMO systems. Firstly, to mitigate the grid modeling errors, we introduce an efficient grid updating algorithm that calculates the roots of a polynomial. Subsequently, to address the loss of sampled data caused by subspace leakage under low SNR and small snapshots, we propose a pseudo-noise resampling technique that reduces estimation errors through appropriate noise introduction followed by multiple resampling iterations. Furthermore, to expedite the convergence of the algorithm, we employ unitary transformation and singular value decomposition (SVD) to reduce data volume. Simulation results demonstrate superior estimation performance of our proposed algorithm compared to existing approaches.