Enhanced Performance with Neural Network Based Hybrid Beamforming in Sparse MIMO Systems

The Millimeter-Wave (mmWave) Multiple-Input Multiple-Output (MIMO) systems offer high data rates and capacity but face challenges in sparse propagation environments. Beamforming is one of the techniques by which these challenges can be addressed. It can be done by searching for an appropriate beam and accurately aligning the beam in the direction of User Equipment (UE). Hybrid Beamforming (HBF) has emerged as a promising solution, combining analog and digital processing to improve performance and by minimizing hardware requirements. Compared to an exhaustive search, the overhead in the beam selection can be reduced by using machine learning and the subset of it, Neural Networks (NN). This paper presents a novel approach to enhance the performance of mmWave MIMO systems by integrating Neural Networks (NNs) with hybrid beamforming. Our proposed Neural Network Hybrid Beamforming (NHBF) method combines multiple streams into a single beam, transmitted via high-order transmission, achieving improved Bit Error Rate (BER) performance compared to traditional Hybrid Beamforming (HBF). By optimizing power distribution, the NHBF beamforming approach eliminates the need for tedious hardware requirements, simplifying the implementation process. Simulation results demonstrate significant performance enhancements, including up to 18% gain in spectral efficiency, a minimum 50% decrease in bit error rate, a 25% increase in energy efficiency, and a 20% reduction in total power consumption compared with Hybrid Beamforming (HBF).