Cloud-Edge based Personalized Federated Learning for In-Home Health Monitoring
Year: 2025
Type: article
Abstract: As the global elderly population grows, there is an increasing need for home-based health monitoring solutions. The integration of Internet of Things (IoT) technologies with machine learning has led to significant advancements in smart healthcare systems. However, most existing methods for in-home health tracking compromise user data privacy, limiting theirscalability and real-world adoption. To address this, we introduce FedHome, an innovative cloud-edge federated learning architecture that enablescollaborative model training while preserving data locality. To manage the challenges of imbalanced and non-IID (non independent and identically distributed) health data, we incorporate a generative convolutional autoencoder (GCAE). Thismodule enhances personalization by generating a class-balanced dataset derived from the user’s own records. Furthermore, GCAE is lightweight, facilitating efficient model exchange between edge devices and the cloud with minimal communication overhead. Comprehensive experiments on real-world human activity datasets confirm that FedHome significantly outperforms conventional approaches in both accuracy and privacy preservation
Author: Leeladhar Revankar, Valencia Veigas
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W6893475631
FWCI:
Cites:
Cited by:
Related to: