225 RALPHI

This research mandate builds upon the existing NEARS-SAPA telemonitoring system to propose a next-generation human activity recognition (HAR) architecture aimed at supporting the autonomy of the elderly at home. Current discrete sensors (PIR and magnetic) restrict detection to basic macroscopic activities and struggle to track multiple individuals simultaneously within a smart environment. To overcome these limitations, this mandate proposes deploying a 3D multi-LiDAR network that fuses spatial data to continuously track multiple people and eliminate blind spots. Furthermore, the system will leverage advanced self-supervised learning AI models to infer fine and complex activities without relying on exhaustive manual data annotation. To train these models, new LiDAR datasets will be constructed in controlled smart environments, capturing specific activities of daily living prioritized in collaboration with clinicians. This approach stands out from current state-of-the-art technologies by providing high clinical granularity while ensuring complete privacy by design, avoiding the ethical issues associated with cameras. Ultimately, this research will position the FRQNT INTER strategic cluster at the international forefront of enabling environments, potentially paving the way for future robotic assistance applications.

Lead applicant