Fall Detection using LiDAR Sensor with PointNet++
Anish Sai Ayyagari, S. A. Ahmed, Shri Venkatakrishnan, Senthil Kumar Thangavel, Gurusamy Jeyakumar, Madhusudana Rao Nalluri · 2025
In today's world, fall incidents pose a significant risk to the elderly, often resulting in severe consequences. Falls can cause damage to the body, especially in elders, where it can be fatal.It is vital to detect falls in order to provide assistance. The simple solution is to install cameras in required areas, which would constantly monitor the area for any falls occurring, this poses a privacy issue in high-risk regions such as bathrooms and toilets. Keeping this in mind, this paper proposes the use of LiDAR sensors that do not cause any invasion of privacy for fall detection. The paper further uses the Pointnet++ to detect fall from the LiDAR scans. The model achieves an accuracy of 85.1 % and precision of 84.8 % on a 70–30 split of the dataset and an accuracy of 86.3 % and precision of 86.2 % on an 80–20 split of the dataset.