Abstract:In this paper, we present a comprehensive study on the application of YOLOv8, a state-of-the-art computer vision (CV) model, to the challenging problem of joint detection and classification of signals in a highly dynamic and congested RF environment. Using our synthetic RF datasets, we explored three different scenarios with congested communication and radar signals. In the first study, we applied YOLOv8 to detect and classify multiple digital modulation signals coexisting within a highly congested and dynamic spectral environment with significant overlap in both frequency and time domains. The trained model was able to achieve an impressive mean average precision (mAP) of 0.888 at an IoU threshold of 50%, signifying its robustness against spectral congestion. The second part of our research focuses on the detection and classification of multiple polyphase pulse radar signals, including Frank code and P1 through P4 codes. We were able to successfully train YOLOv8 to deliver a nearly perfect mAP50 score of 0.995 in a densely populated signal environment, further showcasing its capability in radar signal processing. In the last scenario, we demonstrated that the model can also be applied to the multi-target detection problem for continuous-wave radar. The synthetic datasets used in these experiments reflect a realistic mix of communication and radar signals with varying degrees of interference and congestion - a setup that has been overlooked by many past research efforts, which have primarily focused on ML-based classification of digital communication signal modulation schemes. Our study demonstrated the potential of advanced CV models in addressing spectrum sensing challenges in congested and dynamic RF environments involving both communication and radar signals. We hope our findings will spur further collaborative efforts to tackle the complexities of congested RF spectrum environments.

VOMTC: Vision Objects for Millimeter and Terahertz Communications

Vision-Position Multi-Modal Beam Prediction Using Real Millimeter Wave Datasets

Applying Deep-Learning-Based Computer Vision to Wireless Communications: Methodologies, Opportunities, and Challenges

ViWi Vision-Aided mmWave Beam Tracking: Dataset, Task, and Baseline Solutions

Role of Sensing and Computer Vision in 6G Wireless Communications

DeepSense-V2V: A Vehicle-to-Vehicle Multi-Modal Sensing, Localization, and Communications Dataset

Vehicle Cameras Guide mmWave Beams: Approach and Real-World V2V Demonstration

Multi-View Millimeter-Wave Imaging over Wireless Cellular Network

Advancing Ultra-Reliable 6G: Transformer and Semantic Localization Empowered Robust Beamforming in Millimeter-Wave Communications

Target Detection and Recognition for Wide Area Border Defense Intelligent Surveillance.

Millimeter Wave Drones with Cameras: Computer Vision Aided Wireless Beam Prediction

Design and Implementation of a 2D MIMO OCC System Based on Deep Learning

Beamforming and Scalable Image Processing in Vehicle-to-Vehicle Networks

ViWi: A Deep Learning Dataset Framework for Vision-Aided Wireless Communications

Machine Learning-Based mmWave MIMO Beam Tracking in V2I Scenarios: Algorithms and Datasets

Deep Learning based Computer-vision for Enhanced Beamforming

A Novel Visible-Depth-Thermal Image Dataset of Salient Object Detection for Robotic Visual Perception

Sensing-Aided 6G Drone Communications: Real-World Datasets and Demonstration

Vision-aided Multi-user Beam Tracking for Mmwave Massive MIMO System: Prototyping and Experimental Results

Joint Detection and Classification of Communication and Radar Signals in Congested RF Environments Using YOLOv8

Vision-Aided Beam Tracking: Explore the Proper Use of Camera Images with Deep Learning