Augmenting Beam Alignment for mmWave Communication Systems via Channel Attention

Abstract

The beamforming technique has attracted considerable attention in wireless communication due to its various advantages such as interference reduction and improved wireless resource efficiency. However, the beam alignment between transmitting and receiving devices, which is fundamentally required for the beamforming, poses a significant challenge due to the continuous variability of the wireless channel. Recently, a deep learning-based technique has been proposed to predict narrow beam indices by measuring wide beams. However, there is room for improvement in the performance of the neural network architecture employed in this technique. Therefore, we suggest a novel deep learning model architecture that incorporates a channel attention module for beam training. The simulation results show a significant enhancement in performance with our scheme compared to both a state-of-the-art approach and other existing methods across all scenarios. Particularly, we confirm that even when reducing the number of wide beams used for measurement by approximately 50%, our proposed approach achieves a performance close to that of the state-of-the-art scheme.