Novel Fall Detection Algorithm based on Multi-Threshold Fall Model

Abstract

This paper elucidates an advanced, multi-threshold-based human fall detection algorithm, employing acceleration sensor data to revolutionize fall risk management in high-risk populations such as the elderly and mobility-impaired individuals. The data procured is meticulously analyzed and pre-processed, with various indicators employed in selecting appropriate parameters for data management. A key innovation of this study is the application of multiple thresholds, an enhancement leading to increased accuracy and reliability in distinguishing real falls from non-fall activities. Optimal thresholds were determined using a boxplot, facilitating a more precise fall detection system. Impressively, this approach achieved 95.45% fall detection accuracy, indicating its potential for practical integration. This research substantially contributes to the safety of individuals prone to falls.