SF2D: Semi-supervised Federated Learning for Fall Detection using (Un)labelled Data in Edge-Cloud
Seyed Alireza Rahimi Azghadi, Hung Truong Nguyen Thanh, Irina Kondratova, Hélène Fournier, Mónica Wachowicz, Francis Palma, René (1914-....). Auteur ou responsable intellectuel Richard, Hung Cao · 2025
The aging population faces increased health risks, with falls being a major concern for individuals over 65, leading to healthcare strain and distress. We propose a semi-supervised federated learning-based fall detection (SF2D) method that leverages edge devices to maintain user privacy while ensuring accurate detection. Our approach first trains an unsupervised autoencoder with federated learning, then uses its encoder to train a cloud-based classifier with benchmark datasets. Our proposed SF2D improves accuracy by 1% and recall by 4% over state-of-the-art systems, offering a practical, accurate solution for fall detection and elderly care.