On high resolution bearing estimation for scattered sources

Previous studies in urban wireless communications underscored an angular spread of the signals reaching the base station antenna array. In this case, many of the array processing methods based on the point source assumption produce poor performance. This paper highlights the behavior of subspace-based methods in such scenarios. A generalization of the MUSIC location algorithm to more realistic propagation models is derived, taking into account a parametric distribution of the angular density power. Moreover, this work proposes and motivates the use of linear prediction methods as an alternative for high resolution bearing estimation in presence of angular spreading. Indeed, the deconvolution of an auto-regressive spectrum permits one to obtain both the direction of arrival and the angular spread of the assumed parametric shape. A linear prediction method, optimally applied to a “Lorenz” shape of the angular distribution is also implemented. In fact, it seems that a Lorenz shape matches better experimental results than Gaussian or uniform ones