Multi-Class Fall Detection Based on Machine Learning by using FallAIID dataset

Abstract

The prevalence of falls among older adults is a significant concern for healthcare professionals and researchers. Preventing fall-related injuries and deaths requires accurate and efficient fall detection systems. This paper proposes a novel approach for multi-class fall detection using machine learning techniques with the FallAIID dataset, evaluates its performances, and proposes an efficient low-cost prototype hardware system. The proposed method leverages the unique characteristics of the FallAIID dataset to accurately classify different types of falls and daily activities with an accuracy of 96% for wrist-worn devices and 95% for neck and waist worn devices. The results of our evaluation demonstrate the effectiveness of our approach and its potential to improve fall detection in real-world settings.