Direction-of-Arrival Estimation with Arbitrary Linear Array: A Turbo-HS Approach

Conventional subspace-based direction-of-arrival (DOA) estimation algorithms require optimal environments to achieve satisfactory estimation accuracy. With the advancement of sparse signal recovery theory, sparse optimization-based DOA estimation algorithms have exhibited commendable performance. One of the popular sparse optimization method is sparse Bayesian learning (SBL), due to the fact that it does not require regularization parameters. However, existing SBL algorithms do not adequately address the issue of noise modeling. In order to more accurately model the received signal covariance matrix under finite snapshots, we transform the signal covariance matrix into the form of sparse representation, normalize the variance of the observation matrix and vectorize it. Based on this, we introduce the Turbo-CS framework, which divides the DOA estimation problem into two subproblems: LMMSE estimation for external noise separation and a novel prior for internal SBL DOA estimation. As demonstrated by the experimental results, the proposed algorithm performs optimally in a variety of scenarios.