Low-Complexity Adaptive Beamforming Algorithm With High Dimensional and Small Samples

A large-scale array (LSA) inevitably encounters scenarios with a small number of samples, and its beamformer suffers from high computational complexity. High computational complexity prevents the system from being used in practical online engineering applications. The complex vector of the beamformer weights can be expressed as the product of training snapshots and the signal steering vector (SSV), and a coefficient vector, since the optimal weight vector is a linear combination of basis vectors of the signal-plus-interference subspace. In this study, a new adaptive beamformer is developed based on the minimum variance distortionless response (MVDR) criterion and kernel techniques. The new beamformer only needs to compute the inversion of a low-dimensional Gram matrix instead of the high-dimensional sample covariance matrix (SCM), which significantly reduces the calculation cost. Moreover, an efficient loading parameter calculation method (only related to the received matrix and not required user-defined parameters) is derived, which can adaptively suppress the mismatches of the ill-conditioned Gram matrix. Furthermore, a fast version of the new beamformer is formulated for the LSA under the scanning mode. Simulation results demonstrate that the new beamformer achieves better performance and a lower computation load than existing algorithms for a small number of samples. In particular, insufficient samples and high computational complexity problems are more frequently aroused in space-time broadband array signal processing. Interestingly, the new techniques can be successfully extended to wideband array signal processing and yield satisfactory beam pattern shapes.