Enhanced Convergence Speed For DOA Estimation Utilizing Covariance-Free SBL Algorithm

Sparse Bayesian learning (SBL) is widely applied in direction of arrival (DOA) estimation for its noise robustness and excellent performance. However, it involves high computational costs due to the covariance matrix inversion required in each Expectation-Maximization iteration. In this paper, we propose an efficient DOA estimation method based on Covariance-Free SBL, which converts the matrix inversion issue into a linear equation-solving problem via a diagonal estimation criterion. Then we employ the complex conjugate gradient algorithm to solve the complex symmetric linear equations, for its advantages of fast convergence, memory efficiency, and accuracy controllability. Simulation results indicate that our method significantly enhances calculation speed meanwhile maintaining estimation accuracy, especially for high grid-resolution DOA estimation.