Image Filtering Techniques for Beam Prediction in a Real-world 6G UAV Scenario

• DOI: 10.1109/wsce59557.2023.10365979
• Publication Date: 2023-12-25T19:40:57Z
• Authors: Vasileios P. Rekkas; Sotirios P. Sotir...; Panagiotis Sarigi...; Konstantinos E. P...; George K. Karagia...; Sotirios K. Goudos
• Keywords: Extremely high frequency; RGB color model; Feature (linguistics)

Millimeter-wave (mm-wave) and terahertz (THz) communication systems can satisfy the high data rate requirements in 5G, 6G, beyond networks, but still rely on use of extensive antenna arrays to guarantee sufficient received signal strength. Many antennas incur beam training overhead; thus, narrow beams require adjustment support highly mobile applications. Deep learning (DL) vision-aided solutions potentially forecast optimal beams, leveraging raw RGB images captured at base station. Image filtering techniques have been widely used computer vision (CV) modify enhance quality an image, based specific rules. This work applies different filters for accurate mm-wave/THz prediction feature extraction pre-trained convolutional neural networks (CNNs). The assessment developed framework is conducted actual dataset by unmanned aerial vehicle (UAV) operating millimeter-wave frequency range. comprises taken Ensemble are also studied, enhancing accuracy two state-of-the-art (SOTA) DL models.