Linear prediction methods for scattered sources
T. Abdellatif, P. Larzabal, H. Clergeot · 1999
Signal array processing appears today as a means to improve the current wireless network thanks to a better estimation of the propagation channel's parameters. This paper focuses on the bearing estimation issue. Indeed, location algorithms are usually based on the point source hypothesis, assuming that the antenna is radiated by plane waves. However, previous studies in urban areas highlighted an angular spreading of the impinging rays due to scatterers bringing about diffuse multipaths. In this case, conventional subspace based methods provide disappointing results. This article proposes to study another kind of high resolution methods such as linear prediction ones. Thus, a deconvolution technique is proposed to return both the angular spread and the direction of arrival of so-called distributed sources. This latter method is a model fitting approach. A spectral smoothness prior information is then introduced to permit a more flexible use of the previous algorithms. The simulation results are compared to the maximum likelihood method and to the Cramer Rao bound.