A Computer Vision Based Beamforming Scheme for Millimeter Wave Communication in LOS Scenarios

Abstract

A novel location-aware beamforming scheme for millimeter wave communication is proposed for line of sight (LOS) and low mobility scenarios, in which computer vision is introduced to derive the required position or spatial angular information from the image or video captured by camera(s) co-located with mmWave antenna array at base stations. A wireless coverage model is built to investigate the coverage performance and influence of positioning accuracy achieved by convolutional neural network (CNN) for image processing. In addition, videos could be intentionally blurred, or even low-resolution videos could be directly applied, to protect users' privacy with acceptable positioning precision, lower computation complexity and lower camera cost. It is proved by simulations that the beamforming scheme is practicable and the mainstream CNN we employed is sufficient in both aspects of beam directivity accuracy and processing speed in frame per second.

Computer Vision Aided Beamforming Fused With Limited Feedback

“…TDD systems, on the other hand, are also subject to reciprocity loss, as the different transceiver circuits in TDD systems can lead to asymmetric links, especially if there are calibration errors in large-scale arrays [22], besides the loss from time-varying effect of the channel. So in this paper, we are going to look into this problem from another perspective [5] [11]. Wireless channel could be predicted by RGB images captured by cameras [11][13] [19].…”

“…To address the problems above, computer vision (CV) aided beamforming has become a feasible solution. In our previous work, CV was utilized to select mmWave beams in LOS scenario with no or limited wireless overhead, for mmWave signals have good directivity and similar propagation characteristic as visible light [5]. In our work, visual detection also can save lots of overhead in various scenes, such as electromagnetic exposure control [6] for high-gain arrays and over-the-air phase calibration [7] for phased arrays.…”

Wireless Channel Prediction in Partially Observed Environments

“…There has been growing interest in combining radio frequency (RF) sensing, such as RADAR, with other sensing modalities such as camera or LIDAR [1], [2]. Camera data have also been proposed to guide communications, such as beamforming at millimeter-wave (mmWave) frequencies [3], [4]. A basic problem in these applications is to predict RF propagation from visual information from cameras or other sources.…”