An Intelligent Vision-based System for Real-Time Abnormal Behavior Detection in Healthcare Environments

Abnormal behaviors among patients, such as falls, sudden collapses, or unusual movements, pose serious safety concerns in healthcare settings and often act as early indicators of medical emergencies. Continuous monitoring is essential, yet traditional systems rely heavily on manual supervision or wearable devices that are uncomfortable, costly, or impractical for long-term use. To address these limitations, this work proposes a low-cost, non-invasive abnormal behavior detection system using deep learning. The system integrates USB cameras with a Raspberry Pi running a YOLO-based model for real-time human detection and posture analysis. Upon detecting abnormal activity, alerts are generated through a buzzer, LCD display, and GSM module. The hardware includes onboard memory, communication modules, and audiovisual feedback interfaces. This approach eliminates wearable sensors, reduces caregiver workload, and enables automated monitoring across hospitals, elderly care centers, and home environments. By combining spatial and temporal analysis, the system achieves reliable detection with low power consumption, providing a practical alternative to existing solutions and significantly improving patient safety.