Robust localization of scattered sources

This paper presents a new robust algorithm for scattered source localization. The proposed algorithm is based on a decomposition of the channel vector into subspaces characterized by their sensitivities to the spatial source parameters, such as the source spread which is usually treated as an unknown nuisance parameter. This decomposition isolates a subspace of the data which is not a function of the unknown nuisance parameters, and the resulting estimator does not involve any search over these parameters. A maximum likelihood estimator for the new decomposed model is developed. The estimator uses only the information carried by the insensitive subspace of the data while perturbations of the channel vector in the sensitive subspace are assumed to be unknown parameters. Identification of the insensitive subspace is done according to the channel vector covariance matrix. Simulation results are presented to demonstrate the effectiveness of the proposed algorithm.