Adaptive Beamforming via Sparsity-Based Reconstruction of Covariance Matrix

Abstract

Traditional adaptive beamformers are very sensitive to model mismatch, especially when the training samples for adaptive beamformer design are contaminated by the desired signal. In this chapter, we reconstruct a signal-free interference-plus-noise covariance matrix for adaptive beamformer design. Exploiting the sparsity of sources, the interference covariance matrix can be reconstructed as a weighted sum of the outer products of the interference steering vectors, and the corresponding parameters can be estimated from a sparsityconstrained covariance matrix fitting problem. In contrast to classical compressive sensing and sparse reconstruction techniques, the sparsity-constrained covariance matrix fitting problem can be effectively solved as a modified least squares solution by using the a priori information on the array structure. Extensive simulation results demonstrate that the proposed adaptive beamformer almost always provides the near-optimal output performance regardless of the input signal power.