A Real-Time Vision-based Fall Detection System using YOLOv10 and Deep Learning for Ensuring Elderly and Patient Safety

Falls among elderly individuals and patients with mobility impairments pose a critical challenge in hospital environments, often leading to severe injuries, prolonged hospital stays, and increased healthcare costs. Traditional fall detection methods, such as pressure sensors and infrared monitoring, frequently suffer from high false alarm rates, delayed response times, and patient discomfort, limiting their effectiveness in realworld hospital applications. To address these challenges, this paper presents an AI-driven Smart Bed Exit Alarm System designed specifically for hospital patient monitoring. The system leverages real-time motion tracking and deep learning to detect falls with high precision. Using computer vision techniques, it continuously analyses movement patterns, accurately differentiating routine patient activities from potential falls. Upon detecting an abnormal event, the system immediately activates an audible alarm and sends real-time mobile notifications to hospital caregivers, ensuring rapid medical intervention. Additionally, it records incident footage for post-event analysis, supporting improved patient safety protocols and hospital fall prevention strategies. Experimental evaluations demonstrate a detection accuracy of 97.2%, significantly reducing false positives, making the system a highly reliable, non-intrusive, and efficient hospital-based patient monitoring solution. By integrating automated alerts, continuous surveillance, and AIdriven risk assessment, the proposed system enhances hospital patient safety, accelerates emergency response times, and optimizes healthcare management in medical facilities.