Adaptive beamforming is a method for enhancing desired signals while suppressing noise and interference received by an array of sensors. Traditional adaptive beamforming techniques have been designed for point sources which travel along a single path to an antenna array. However, in practical applications such as wireless communications, multipath propagation due to local scattering of the sources causes spreading of the signal energy around the nominal direction-of-arrival (DOA). We consider the problem of adaptive beamforming for enhancing sources with uncertain spatial distributions.

Using a Bayesian approach, the source DOA is assumed to be a discrete random variable, and the beamformer which minimizes the mean square estimation error is shown to be a weighted sum of spatial Wiener filters pointed at a set of candidate DOAs. The relative contribution of each Wiener filter is determined from the a posteriori distribution of the DOA, which functions as a non-parametric estimate of the spatial distribution. Simulation examples demonstrate the performance of the proposed technique for a variety of multi-user scenarios using different spreading models.

This is a report of joint work with Harry L. Van Trees, Yariv Ephraim, and Lillian Xiaolan Xu