Generalized Framework to Subspace-based DOA Estimation

A generalized framework to subspace-based direction-of-arrival (DOA) estimation in the presence of unknown noise field and other imperfections is proposed. Provided the noise covariance matrix estimate, a forward-only and its forward-backward extension rooting-based DOA estimator is developed. The DOA estimator takes advantage of using second-order statistics of signal subspace perturbations in constructing a weight matrix of a properly designed generalized least squares minimization problem. Only few iterations are sufficient to reach high accuracy by the proposed DOA estimator. The consistency of the estimator is also proven. Then a DOA selection strategy with low computational cost is developed to select actual sources out of larger (double) number of candidates generated using the proposed DOA estimator. Numerical study is conducted in order to establish the significant superiority of the proposed generalized framework for subspace-based DOA estimation over the existing state-of-the-art methods, especially in challenging scenarios with imperfections such as different noise covariance structures, small sample size, low signal-to-noise ratio, correlated and/or closely located sources, as well as very large number of sources.