Sparse Array Wideband Beamforming Based on Covariance Matrix Reconstruction and Neural Network

Sparse arrays offer economic advantages by reducing the number of antennas. However, directly utilizing the covariance matrix of sparse array signals for wideband beamforming may lead to the emergence of grating lobes, seriously impairing the performance of massive multiple-input-multiple-output (MIMO) systems. In this paper, a sparse array wideband digital beamforming algorithm based on the difference coarray (SWDBF-DCA) is proposed. Utilizing the correspondence between the difference coarray and the path difference of the received signals, the signal covariance matrix obtained from the sparse array is reconstructed. Subsequently, the reconstructed covariance matrix is employed for adaptive wideband beamforming, effectively addressing the grating lobe issue caused by antenna spacing larger than half-wavelength. Furthermore, to address the high computational complexity associated with covariance matrix inversion in the aforementioned algorithm, a sparse array wideband digital beamformer based on the neural network (SWDBF-NN) is proposed. Leveraging the fitting capability of neural networks, a nonlinear mapping between the input sparse signals and the output beamforming weights is constructed. The utilization of a complex network structure preserves the phase information of complex signals, enabling effective beamforming for sparse arrays with limited snapshots while reducing computational complexity. Simulation results verify the effectiveness of the two proposed algorithms.