Abstract

Accidents pose a significant threat worldwide, often leading to severe harm and loss. Existing solutions mainly focus on vehicle-related accidents and rely heavily on smartphones, leaving a gap in detecting and alerting accidents outside vehicular contexts. This study proposes an IoT- and machine learning-based personalized accident detection and tracing system to address this limitation. The system comprises an IoT-enabled smart band equipped with sensors to monitor vital signs (heart rate, blood pressure, body temperature, and SpO₂) and GPS for precise location tracking, a user-specific machine learning model to identify abnormal physiological states, and a cross-platform mobile application to deliver real-time emergency alerts and location information to responders. Sensor readings are transmitted via Wi-Fi to a cloud server, minimizing smartphone dependency and latency compared to GSM/GPRS-based systems. The ML model, trained on both public and locally collected datasets, achieved 99.44% accuracy using a Random Forest classifier. Validation against medical-grade devices showed strong measurement agreement (Pearson correlation > 0.94), while field trials confirmed stable device operation and cross-platform compatibility. Additionally, the research contributed a large, locally obtained physiological dataset valuable for future accident prevention and healthcare research. By reducing detection-to-response time and enhancing predictive accuracy, the system offers a significant advancement in safeguarding individuals from common accidents.

Statistics