Novel beamspace neural network approach to mobile unit localization

A novel beamspace radial basis function neural network for the estimation of the angle-of-arrival (AOA) of a mobile unit in cellular communications is proposed. By training the network with the data emitted from different sub-cells in the field of interest, and then collected by a set of antenna array beamformers optimum weights which lead to the best approximation of the desired response in least-square sense can be obtained. In principle, the network performs mapping from the complex input data into the desired angle response. Computer simulations demonstrate that the proposed scheme is effective in combating multipath interference.