An Edge-Cloud Collaborative Architecture for Proactive Elderly Care: Real-Time Risk Assessment and Three-Level Emergency Response
A new AI architecture fuses data from five sensor types to generate emergency alerts in under 3 seconds.
Researchers Lijie Zhou and Luran Wang have published a paper detailing a new edge-cloud collaborative architecture designed to solve critical shortcomings in existing elderly care monitoring. Current cloud-centric systems suffer from high latency, privacy risks from constant data transmission, and simplistic alerting. Their proposed framework uses a five-layer design (device, edge, service, data, application) to enable real-time processing. At its core, a weighted multi-modal fusion algorithm running on edge devices like Raspberry Pi 4 gateways integrates data from five different sensor types, achieving sub-100 millisecond inference latency and preserving privacy by processing raw data locally instead of sending it to the cloud.
The system generates a unified risk score by analyzing four dimensions: fall probability, physiological indicators, behavioral patterns, and sensor anomalies. Based on dynamic thresholds, it activates a sophisticated three-level emergency response, automatically coordinating notifications between family members, community doctors, and nearby volunteers. Experiments on standard datasets (CASAS, MIMIC-III, SisFall) show the approach outperforms single-sensor methods, achieving 91% activity recognition accuracy and an 84% F1-score for anomaly detection. Most critically, the entire pipeline from sensor detection to alert delivery operates with an end-to-end latency of under three seconds, making it viable for genuine emergency intervention.
- Uses a five-layer edge-cloud architecture to process data locally on Raspberry Pi 4, achieving sub-100ms inference and keeping raw data private.
- Fuses data from five sensor types via a weighted algorithm, scoring risk across four dimensions (falls, physiology, behavior, anomalies) with 91% recognition accuracy.
- Triggers a three-level alert system (family, doctor, volunteer) with end-to-end latency under 3 seconds, a 84% F1-score for anomaly detection.
Why It Matters
This architecture provides a practical, privacy-preserving blueprint for scalable, real-time in-home health monitoring that can save lives.