Robust auto-focusing wideband DOA estimation

Within the last decade there has been a growing interest in developing techniques for the estimation of the direction of arrival (DOA) of waveftonts carrying wideband signals in order to locate the emitting sources. The coherent signal-subspace method (CSM) is one of the most largely adopted iterative technique based upon the concept of signal-subspace, due to its ability in handling coherent sources, while showing very good detection and resolution thresholds, low bias and high accuracy. Central to CSM is the use of the so-called focusing matrices, whose characteristics strongly influence its overall performance. In the literature, several focusing matrices have been proposed and the most effective of them often require initial estimation of the DOAs, which could be a drawback for the overall estimation procedure. Furthermore classical focusing design techniques do not take into account the fact that at each iteration of the algorithm estimated DOAs may differ from the actual ones on which the manifold should be focused. In this paper we propose a novel focusing matrices design technique aimed at counteracting some of the main disadvantages of other classical focusing matrices. It is shown that the classes of Rotational Signal Subspace and Signal Subspace Transformation focusing matrices leave room for further optimization of the available degrees of freedom, that are exploited here to create a CSM robust against DOA estimation errors. For this reason, the proposed method is referred to as robust coherent signal-subspace inethod (R-CSM). Thanks to the peculiarities of these novel focusing matrices, the initial preprocessing stage of the classical CSM is no longer required. Furthermore, the convergence speed is improved, because at each iteration the degrees of freedom are used to concentrate the focusing closer and closer about the estimated DOAs. (c) 2005 Elsevier B.V. All rights reserved.