Wi-Fi CSI-Based HAR for Non-Intrusive Monitoring of the Elderly

CONTEXT The aging population necessitates the development of non-intrusive and privacy-respectful home monitoring solutions. Wi-Fi CSI (Channel State Information) signals offer a promising alternative to cameras and wearable sensors. This internship aims to design a Human Activity Recognition (HAR) system dedicated to monitoring the elderly via Wi-Fi CSI. INTERNSHIP OBJECTIVES • Recognition of daily activities (walking, sitting, standing, falling) • Robustness to environments (different rooms, furniture, conditions) • Detection of behavioral anomalies (falling, prolonged inactivity) TASKS REQUIRED • Bibliographic review of HAR based on Wi-Fi CSI • Setup of the CSI data collection and preprocessing pipeline • Development of the recognition model (CNN, Transformer) • Domain adaptation for multi-environment robustness • Field validation in a realistic environment DELIVERABLES • Operational CSI collection and preprocessing pipeline • Trained and validated HAR model • Anomaly detection module • Scientific report