Lightweight Cluster Matching-Based Tracking Method for Fall Detection Using a Single-LiDAR

Abstract

Falls are a major health risk in super-aged societies, particularly for elderly individuals living alone. Light detection and ranging (LiDAR) sensors have recently attracted attention for fall detection due to their robustness to lighting conditions and inherent privacy preservation. However, practical in-home deployment requires continuous operation under limited computational resources. This article proposes a lightweight fall detection method using a single-LiDAR sensor, integrating cluster matching-based tracking with rule-based acceleration analysis. The proposed approach enables continuous monitoring without multisensor configurations or learning-based models. Experimental results across diverse fall and nonfall scenarios demonstrate an accuracy of 89.2% and a sensitivity of 98.3% while requiring only 1.02 MFLOPs per frame. The results confirm the feasibility of reliable and computationally efficient fall detection using a single-LiDAR sensor.