Bias-Compensated MPDR beamformer for small number of samples

Abstract

Adaptive beamforming is a central processing stage in many sensor array applications. Minimum Power Distortionless Response is one of the most popular technique, but suffers from strong degradation when the sample covariance matrix is ill-conditioned due to small sample support. Many robust beamformers have been designed to circumvent this drawback, such as diagonal loading or reduced rank techniques, to cite a few. In this communication we present a new robust beamformer, based on bias analysis of the sample covariance matrix eigenvectors. This beamformer can be viewed as a bias-compensated reduced rank beamformer. This beamformer is shown to have a better behaviour than a principal component beamformer in the case of a weak signal of interest.