Linear Prediction-Based Covariance Matrix Reconstruction for Robust Adaptive Beamforming

In this letter, a novel reconstruction-based adaptive beamformer is proposed, which uses linear prediction to generate virtual sensor data and extend array aperture. To overcome suppression failure of reconstruction-based adaptive beamformer, a double-side array extending algorithm is proposed for uniform linear array, where the virtual array data can be obtained by using the data from real array. Then, we estimate the spatial spectrum of extended array data with higher resolution by using a modified diagonal loading-type procedure, which is decided by the condition numbers of the sample covariance matrix (SCM) and the extended SCM, and the interference-plus-noise covariance matrix (INCM) of extended array is estimated. Without any optimization procedure, steering vector (SV) of extended array is corrected as the eigenvector corresponding to the dominant eigenvalue of the extended SCM. Numerical Simulations demonstrated that the proposed beamformer can use both deep nulls and low side-lobes to ensure better interference suppression capability to other beamformers in the case of sensor position errors.

Automated summary

A novel reconstruction-based adaptive beamformer is proposed utilizing linear prediction to generate virtual sensor data and extend array aperture. To overcome suppression failure, a double-side array extending algorithm is introduced along with methods for estimating the spatial spectrum of extended array data. The proposed beamformer shows better interference suppression capabilities compared to other beamformers in cases of sensor position errors.