Robust Aligned Widely Linear Beamforming

The problem of wideband beamforming for the extraction of a wideband Signal-of-Interest (SOI) contaminated by independent wideband interferences and sensor noise is addressed in the paper. With the exploitation of the non-circularity of wideband SOI/interferences, the aligned widely linear beamformers (AWLBs) in theory are potentially more attractive than the conventional linear wideband beamformers in protecting a rectilinear or almost rectilinear SOI while suppressing rectilinear or non-rectilinear interferences. However, the presence of even a minor look direction error would severely degrade the performance of AWLBs. Two robust AWLBs are proposed in the paper, with and without preliminary SOI non-circularity parameter estimation. The aligned and augmented interference-noise covariance matrix required in the implementation of the proposed beamformers is estimated by using a newly developed sub-band association scheme under finite data samples. Simulation results show that the presented beamformers are robust against the look direction error in SOI estimation and interference-noise suppression.