Deep Learning-Based Hybrid Beamforming Under Impulsive Noise for mmWave MIMO Systems

Abstract

A novel deep learning framework has been proposed for designing hybrid beamformers for a single-user millimeter-wave massive multiple-input multiple-output system that is affected by impulsive noise. Regarding this, we treat the hybrid beamforming design as a regression problem. A Convolutional Neural Network (CNN) model has been designed in this study, which inputs the channel matrix and outputs the hybrid beamformers. The results of the simulation illustrated that the CNN framework successively estimates the best hybrid beamformers and significantly outperforms traditional methods with regard to bit error rate under both impulsive and Gaussian noise.