Optimised statistical performance in an adaptive beamformer

A modified least squares adaptive steered beamformer that aims to eliminate the excess weight noise induced by a strong desired signal is proposed. It functions by alternating projections of the array data onto a subspace orthogonal to both the desired signal and interferences. The desired signal can be finally recovered, while retaining interference cancellation, by the use of a commutation property of the projections. The new method generalizes previous solutions to this problem which have relied on ideal plane wave signal models and it can operate with imperfect knowledge of the array manifold while still achieving the appropriate Cramér-Rao lower bounds on signal estimation.