Elderly Fall Detection Based on YOLO and Pose Estimation

Zhihan Ye

2024

Abstract

Aiming at the increasing risk of falls in the elderly, a fall detection method based on You Only Look Once (YOLO) and Pose Estimation is proposed. The wearable and non-wearable fall detection methods are reviewed. This paper selects a fast and accurate target detection algorithm YOLO. The fall detection data set was used to compare different versions of YOLO (YOLOv5, YOLOv6 and YOLOv8), and finally the accuracy and speed of YOLOv8 were selected. In order to distinguish between falling and lying down, YOLOv8 and Pose Estimation (YOLOv8-Pose) are combined to track key points and motion patterns, achieving a real-time fall detection accuracy of 92%. This method provides a reliable solution for elderly health monitoring.

Paper Citation

in Harvard Style

Ye Z. (2024). Elderly Fall Detection Based on YOLO and Pose Estimation. In Proceedings of the 2nd International Conference on Data Analysis and Machine Learning - Volume 1: DAML; ISBN 978-989-758-754-2, SciTePress, pages 10-17. DOI: 10.5220/0013486500004619

in Bibtex Style

@conference{daml24,
author={Zhihan Ye},
title={Elderly Fall Detection Based on YOLO and Pose Estimation},
booktitle={Proceedings of the 2nd International Conference on Data Analysis and Machine Learning - Volume 1: DAML},
year={2024},
pages={10-17},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013486500004619},
isbn={978-989-758-754-2},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 2nd International Conference on Data Analysis and Machine Learning - Volume 1: DAML
TI - Elderly Fall Detection Based on YOLO and Pose Estimation
SN - 978-989-758-754-2
AU - Ye Z.
PY - 2024
SP - 10
EP - 17
DO - 10.5220/0013486500004619
PB - SciTePress