A deconvolution method for the characterization of distributed sources via linear prediction

This paper deals with high resolution bearing estimation in urban radiocommunication scenarii. Indeed, in such environments, scatterers local to the emitter engender diffuse paths that deteriorate the performances of conventional subspace-based algorithms. A deconvolution technique, involving Linear Prediction methods, is designed to characterize so called distributed sources by returning the mean angle and the angular spreading of the signal angular power density. Two ways of implementation are proposed in two extreme cases of diffuse paths correlations. Simulation results show that this proposed method provides satisfaying results compared to the Cramer Rao-Bound and moreover outperform more famous subspace-based algorithms.