Real-Time Fall Detection Using MLP, OpenCV, and IoT Integration: Development of the FallNex Smart Application

Abstract

Falls among older adults pose a serious health risk, as delayed assistance can lead to severe injuries or fatal outcomes. To address this challenge, we propose FallNex, a real-time fall detection system that integrates a machine learning model, lightweight visual monitoring, and IoT-based hardware. A Multi-Layer Perceptron (MLP) model performs the primary fall prediction using motion data collected from wearable sensors. To minimize false alarms, a simple OpenCV-based monitoring module evaluates frame differences from a camera feed to verify whether significant activity occurred at the moment of a suspected fall. The proposed hardware unit communicates with the FallNex mobile application to deliver instant alerts, maintain fall logs, and provide post-fall guidance through an integrated chatbot. Experimental results demonstrate that FallNex achieves high accuracy with reduced false positives while maintaining real-time responsiveness, making it suitable for home-based and assisted-living environments.