Manuscript received October 31, 2024; revised December 25, 2024; accepted February 14, 2025, published July 17, 2025.

Abstract—In 2020, it was recorded that 5% of the total population in Indonesia were people with disabilities. Individuals with physical disabilities, especially those who cannot use both their hands and feet are facing problem in their mobility. Since most wheelchairs are controlled using hands, these individuals are unable to operate a wheelchair independently. This research aims to create smart wheelchairs that use object detection models, to enable the user to navigate their wheelchair. The smart wheelchair is equipped with a camera that will capture the user’s head movement and will move based on it. Deep learning model algorithms are used to detect the head movement. In this research, three generations of the YOLO (You Only Look Once) model—YOLOv5, YOLOv6, and YOLOv7—are compared to determine the most suitable model for the system. It is found that YOLOv6N has the fastest inference time, that is 2.54 ms. All the models are also evaluated on several parameters: Precision, recall, mAP@.5, and mAP@.5:.95. There’s no huge difference between each variation. All of the precision, recall and mAP@.5 of each variation are above 0.9. Yet, the difference can be seen for the mAP@.5:.95 where the highest score is 0.808 from YOLOv6L and the lowest is 0.703 from YOLOv5N.

Keywords—deep learning, YOLO, head motion, detection, wheelchair, computer vision

Cite: Fitri Utaminingrum, Erlinda Butarbutar, Rahma Nur Fitriyani, Yvette Celia Aviani, and Atik Qomariyah, "Enhancing Smart Wheelchair Navigation through Head Motion Detection: A YOLO-based Approach," Journal of Image and Graphics, Vol. 13, No. 4, pp. 325-335, 2025.

Copyright © 2025 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC-BY-4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.