ROBUST ADAPTIVE BEAMFORMER WITH FEASIBILITY CONSTRAINT ON THE STEERING VECTOR

Wenyi Zhang,Bhaskar D. Rao
DOI: https://doi.org/10.5281/zenodo.39962
2006-01-01
Abstract:The standard MVDR beamformer has high resolution and in- terference rejection capability when the array steering vec- tor is accurately known. However, it is known to degrade if steering vector error exists. Motivated by recent work in ro- bust adaptive beamforming, we develop variants of the con- strained robust adaptive beamformer that attempt to limit the search in the underlying optimization problem to a feasible set of steering vectors thereby achieving improved perfor- mance. The robustness against steering vector error is pro- vided through a spherical uncertainty set constraint, while a set of magnitude constraints is enforced on each element of the steering vector to better constrain the search in the space of feasible steering vectors. By appropriately changing the variables, the optimization problem is modified such that the need for the magnitude constraints are avoided. The devel- oped algorithm is tested in the context of speech enhance- ment using a microphone array and shown to be superior to existing algorithms.
What problem does this paper attempt to address?