Fusion Vision RGB + Wi-Fi CSI for Robust Fall Detection

CONTEXT Fall detection is a critical issue for the safety of elderly individuals at home. Unimodal approaches (vision only or CSI only) have limitations: occlusions for vision, low resolution for CSI. This internship proposes a multimodal approach combining RGB vision and Wi-Fi CSI for robust and reliable detection. INTERNSHIP OBJECTIVES • Reliable fall detection in real-world environments • Tolerance to sensor failures (degraded mode) • Intelligent cross-modal fusion (vision + CSI) • 2D localization of the fall in the room TASKS REQUIRED • Constitution of the multimodal dataset (synchronized RGB + CSI) • Multimodal preprocessing (pose detection, CSI extraction) • Development of the RMFD (Robust Multimodal Fall Detection) model • Implementation of confidence scores by modality • Validation under real conditions DELIVERABLES • Annotated multimodal dataset (RGB + CSI) • Operational RMFD model • Real-time detection system • Scientific report