Robust beamforming via alternating iteratively estimating the steering vector and interference-plus-noise covariance matrix

Abstract

To develop an adaptive beamformer against the steering vector mismatch of the signal of interest (SOI), a novel robust algorithm is proposed to estimate the steering vector of the SOI and interference- plus-noise covariance matrix (INCM) in an alternative and iterative way. That is, via determining a convex optimization problem, which forces the steering vector moving towards the signal-plus- interference subspace (SIS) but getting away from the interference subspace (IS), the actual steering vector of the SOI is estimated. To proceed, the suitable SIS is easy to obtain through applying eigendecomposition on the sample covariance matrix while the appropriate IS is hard to estimate because of the array perturbations. Given this, a novel INCM reconstruction method, which utilizes a blocking matrix to eliminate the SOI from the training samples, is provided to realize the preferable estimate of the IS. More specifically, the abovementioned processes are carried out in an alternative iteration scheme, which leads to the SOI steering vector and INCM converging to the theoretical ones sufficiently, respectively. Unlike the conventional algorithms which are vulnerable to the various mismatches, the proposed beamforming algorithm is insensitive to the SOI steering vector mismatch arisen from the DOA error and array perturbations, numerous theoretical analysis and simulation experiments are presented to demonstrate the superiority of the proposed adaptive beamformer.