Robust adaptive beamforming under data dependent constraints

Abstract

In the conventional robust adaptive beamformers (RABs), a common problem is that the estimated manifold vector of the signal of interest (SOI) may converge to the interference subspace. By using the data dependent constraints, such problem can be mitigated. In this paper, we propose to employ the oblique projection to obtain the SOI covariance matrix and compute the interference covariance matrix in an estimate-and-subtract manner. Rather than using the semidefinite programming relaxation, we find that the proposed RAB, which belongs to the well known non-convex quadratically constrained quadratic programming problem, can be solved efficiently by the Lagrange multiplier methodology coupled with the Newton’s method. Our simulation results demonstrate the superiority of the proposed method over other previously developed RAB techniques.