Event-Based Real-Time Fall Detection Using YOLOv8 and LSTM
Weihao Liang, Y ongsheng · 2024
Addressing the critical issue of fall detection among the elderly as global aging intensifies, this study develops an innovative system that integrates event cameras with advanced object detection and Long Short-Term Memory Networks (LSTM) to accurately identify fall incidents. The research methodology involves the utilization of these technologies to capture dynamic changes in the environment and analyze data through machine learning algorithms, ensuring the system operates effectively even in low-light and privacy-sensitive environments. The experiments conducted on the URFall and Le2i datasets demonstrate a 100%accuracy rate in fall detection, proving the system's efficacy in real-world scenarios. These findings indicate that the proposed system not only significantly improves the safety of elderly populations by providing timely and accurate fall detection but also pushes forward the boundaries of computer vision technology in health care applications.