An AI-driven wearable health device framework achieved 96.1% classification accuracy with 30 ms latency, reduced power consumption by 50%, and reached 98.9% tampering detection.
Benchmark biosignal datasets (PhysioNet, MIMIC-III) and embedded wearable health device (WHD) prototypes
AI-driven, four-tier WHD framework integrating biosensors, deep learning models (CNN, LSTM, Transformer), reinforcement learning-based health-aware control (HAC), federated learning, and blockchain-enhanced secure communication
Classification accuracy, latency, power consumption, tampering detection, and privacy risk reduction
An AI-driven wearable health device framework integrating deep learning, reinforcement learning, and federated learning demonstrated high classification accuracy, reduced power consumption, and robust security in simulation and prototype testing.
The integration of health-aware control (HAC) and prognostics within wearable health devices (WHDs) presents a transformative opportunity for personalized healthcare. This study introduces an AI-driven, four-tier WHD framework for real-time physiological monitoring, predictive analytics, and adaptive system control. The architecture integrates biosensors, deep learning models (CNN, LSTM, Transformer), reinforcement learning–based HAC, federated learning for decentralized intelligence, and blockchain-enhanced secure communication. Experimental validation was conducted using benchmark biosignal datasets (PhysioNet, MIMIC-III) and embedded WHD prototypes, while OMNeT++ simulations evaluated large-scale deployments. Transformer models achieved 96.1% classification accuracy with 30 ms latency; RL-based adaptive sampling reduced power consumption by 50%; and the federated security framework reached 98.9% tampering detection with 90% privacy risk reduction. Wi-Fi consistently outperformed BLE in latency and scalability. These results confirm the feasibility of integrating HAC and prognostics in WHDs, enabling proactive monitoring, energy-efficient operation, and privacy-preserving personalized healthcare.
Building similarity graph...
Analyzing shared references across papers
Loading...
Alok Jain
Pandit Deendayal Petroleum University
Suman Bhullar
Thapar Institute of Engineering & Technology
Array
Thapar Institute of Engineering & Technology
Pandit Deendayal Petroleum University
Building similarity graph...
Analyzing shared references across papers
Loading...
Jain et al. (Thu,) reported a other. AI-driven wearable health device framework was evaluated on Classification accuracy, latency, power consumption, and tampering detection. An AI-driven wearable health device framework achieved 96.1% classification accuracy with 30 ms latency, reduced power consumption by 50%, and reached 98.9% tampering detection.
synapsesocial.com/papers/6a19ce657081f56b37df1ba0 — DOI: https://doi.org/10.1016/j.array.2025.100532