Robust Adaptive Beamforming Based on Sparse Bayesian Learning and Covariance Matrix Reconstruction

Sparse Bayesian learning (SBL) exploits sparse information from a Bayesian perspective by making sparse prior assumptions about the signal. It has good flexibility in modeling sparse signals and can also achieve good estimation results when the number of samples is low. In this letter, we propose a robust adaptive beamforming (RAB) method based on interference-plus-noise covariance matrix (INCM) reconstruction using SBL. This method derives the fixed point iteration formula of the signal under the Laplace prior distribution by maximizing the marginal likelihood function of the data received by the array. Then the sampling position of the coarse grid in the SBL model is regarded as an adjustable parameter, and it is refined through iteration. By using the superiority of SBL in modeling sparse signals, the steering vectors and power of the source signals are precisely estimated, and an accurate INCM is reconstructed to obtain a high output SINR. The only prior knowledge required is imprecise angular sector knowledge of the desired signal (DS). Simulation results demonstrate that our proposed method achieves near-optimal output SINR over a wide range of input SNR and a limited number of snapshots and even achieves good performance in the single snapshot situation.