In this paper, we present a novel radial basis function network (RBFN) assisted wide-band beamformer which is based on a derivation of the Bayesian detector. The proposed receiver structure significantly outperforms conventional linear wide-band beamformers in terms of the achievable angular resolution and thus the receiver is capable of supporting a higher number of users. For the further enhancement of the BER performance, and for the sake of complexity reduction, a decision feedback aided RBF scheme is proposed.