Adaptive fourth-order tensor beamformer

Abstract

A novel tensor-based beamforming algorithm via multi-translation-invariant vector-sensor array is proposed. Largely different from the traditional ones, the spatial filtering is carried out on both macroscopic scale (between subarrays) and microscopic scale (between sensors of each subarray) in the new algorithm, in order to obtain the beamformer output. With different weight vectors being used in each scale, performance dominance of diverse algorithms can be combined effectively. Moreover, the contraction of covariance tensor implies smoothing operation to reduce the singularity of dual-scale covariance matrices. Consequently, the robustness of proposed algorithm to look direction and element position mismatch is increased in the case of high input-SNR (Signal-to-Noise-Ratio). Theoretical analysis and numerical simulations indicate that Dual-Scale-Combined (DSC) algorithm outperforms traditional beamformer in terms of robustness and convergence rate.