Robust Adaptive Beamforming Based on Sparse Representation and Blocking Matrix Construction

Abstract

Adaptive beamformer is susceptible to model mismatch, extraordinarily when the signal of interest (SOI) resides in array observation data. Different from the existing robust adaptive beamforming (RAB) based on the reconstruction of interference-plus-noise covariance matrix (IPNCM), this article introduces sparse representation theory as a means of removing noise from the observation data. This is followed by eliminating the SOI component through the construction of the SOI blocking matrix. Consequently, a relatively pure interference signal can be obtained, which effectively suppresses the unexpected components, namely, the cross-covariance matrix between noise, interference signals and the SOI, in subsequent higher order statistical calculations. Reconstruction of the IPNCM can be accomplished by simply summing the interference covariance matrix with the estimated one of noise. The algorithm's robustness to various model mismatches is further reinforced through the correction of the steering vector, which is implemented by maximizing the SOI power estimator. The simulation results corroborate the efficacy of the proposed method, which is capable of attaining the close-optimal performance and exceeds other methods in the case of multiple steering vector mismatches.