An Edge-Cloud Collaborative Architecture for Proactive Elderly Care: Real-Time Risk Assessment and Three-Level Emergency Response

Lijie Zhou; Luran Wang

eess.SPMar 20, 2026

An Edge-Cloud Collaborative Architecture for Proactive Elderly Care: Real-Time Risk Assessment and Three-Level Emergency Response

Authors: Lijie Zhou, Luran Wang

Organizations: School of Computer Science, University of Nottingham Ningbo China, Ningbo, China · School of Mathematical Sciences, University of Nottingham Ningbo China, Ningbo, China

Abstract

The rapid aging of global populations has created an urgent need for intelligent healthcare monitoring systems to ensure the safety of elderly individuals living independently. Existing cloud-centric platforms face critical limitations, including high latency unsuitable for emergency response, privacy risks from continuous transmission of sensitive data, and limited, single-channel alert mechanisms lacking scalability and context awareness. This paper proposes an edge-cloud collaborative architecture that addresses these challenges through real-time multi-modal sensor fusion, a four-dimensional risk assessment model, and a three-level emergency response system. The framework adopts a five-layer design - device, edge, service, data, and application layers - enabling real-time risk evaluation with end-to-end alert latency under three seconds. At the edge, a weighted multi-modal fusion algorithm integrates data from five sensor types with confidence propagation. A unified risk score is generated by combining fall probability, physiological indicators, behavioral patterns, and sensor anomaly metrics. Based on dynamic thresholds, a three-tier notification system coordinates responses among family members, community doctors, and nearby volunteers. Experiments on CASAS, MIMIC-III, and SisFall datasets show that the approach achieves 91% activity recognition accuracy and an 84% anomaly detection F1-score, outperforming single-sensor methods. Deployment on Raspberry Pi 4 gateways demonstrates sub-100 ms inference latency while preserving privacy by keeping raw data local. This architecture advances practical, privacy-preserving, and responsive elderly care systems.

Figures & tables

Fig. 1: Five-layer edge-cloud collaborative architecture showing data flow from physical sensors through edge gateway to cloud services and end-user applications.

Sensor Type	Protocol	Rate	Weight
Smart Wristband	BLE	1–100 Hz	1.0
Motion Sensor	Zigbee	10 Hz	0.8
Camera	WiFi	15–30 Hz	0.9
Door Sensor	Zigbee	Event	–
Bed Sensor	Zigbee	1 Hz	0.6

TABLE I: Sensor Types and Reliability Weights

Fig. 2: Edge gateway processing pipeline showing the flow from sensor data acquisition through fusion, inference, and alert generation.

Fig. 3: Multi-database architecture showing PostgreSQL for business data, InfluxDB for time-series health data, and Redis for caching.

Fig. 4: Four-dimensional risk scoring model showing the integration of fall probability, health risk, behavioral indicators, and sensor anomaly scores with dynamic adjustment factors.

Fig. 5: Dynamic threshold-based alert level determination showing the four alert levels and their corresponding response mechanisms.

Fig. 6: Three-level notification mechanism showing the escalation from family notification through doctor notification to volunteer dispatch based on alert level.

Fig. 7: Web dashboard interface showing real-time health monitoring, alert management, device status, and health report generation capabilities.

Dataset	Subjects	Duration	Samples	Features
SisFall	23	–	17,050	Accel., Gyro (3-axis each)
CASAS (aruba)	1	214 days	2,341,970	Motion, door, temp sensors
CASAS (cairo)	1	27 days	261,728	Motion, door, item sensors
MIMIC-III	40,000+	Varied	Millions	HR, SpO 2 , BP, Resp.

TABLE II: Dataset Statistics

Method	Activity Acc.	Posture Acc.	Anomaly F1
Wristband only	0.82	0.78	0.71
Motion sensor only	0.74	0.65	0.58
Camera only	0.79	0.88	0.63
Equal-weight fusion	0.88	0.86	0.79
Weighted fusion (ours)	0.91	0.89	0.84

TABLE III: Sensor Fusion Performance Comparison

Configuration	Precision	Recall	F1
Full 4D model	0.89	0.93	0.91
Fall only (w=1.0)	0.82	0.79	0.80
Fall + Health (equal)	0.86	0.88	0.87
w/o Trend adjustment	0.87	0.90	0.88
w/o Dynamic adjustment	0.85	0.91	0.88

TABLE IV: Risk Model Ablation Study

Fig. 8: ROC curve for the four-dimensional risk scoring model showing AUC of 0.94.

Stage	Latency
Sensor $\rightarrow$ Edge (BLE)	$\sim$ 30 ms
Time Alignment + Fusion	$\sim$ 15 ms
AI Inference (DL mode)	$<$ 100 ms
AI Inference (Rule mode)	$<$ 5 ms
Risk Assessment	$<$ 10 ms
Notification Delivery (SMS)	$\sim$ 1.5 s

TABLE V: End-to-End Latency Breakdown

Metric	Result
Concurrent WebSocket connections	12,000+
API throughput (requests/second)	3,500
Edge gateway uptime (30-day test)	99.7%
Alert delivery success rate	98.5%
Mean time to first notification	2.1 s

TABLE VI: System Scalability Metrics

Feature	Ours	CarePredict	Med. Guardian	Bay Alarm
Edge AI processing	Yes	No	No	No
Multi-modal fusion	5 sensors	1 sensor	1 sensor	1 sensor
Alert levels	3	Single	Single	2
Privacy (edge-first)	Yes	No	No	No
Behavior analysis	Yes	Limited	No	No
E2E latency	$<$ 3 s	$>$ 10 s	$>$ 15 s	$>$ 8 s

TABLE VII: Comparison with Commercial Systems

Healthcare

Healthcare AI develops and evaluates models and systems for clinical care, medical research, public health, and healthcare operations.

See all 617 papers

A Multi-Modal CNN-LSTM Framework with Multi-Head Attention and Focal Loss for Real-Time Elderly Fall Detection
Mar 20, 2026Lijie Zhou, Luran WangFall Detection Wearable Sensing
School of Computer Science, University of Nottingham Ningbo China, Ningbo, China · School of Mathematical Sciences, University of Nottingham Ningbo China, Ningbo, China
A Cloud-Edge System for Multimodal Clinical Screening in Resource-Constrained Rural Settings
Aug 13, 2026Hei Ting, Chan, Chenwei Wu +8Efficient Multimodal Inference Healthcare
Una · University of Michigan, Ann Arbor, MI, USA · Massachusetts Institute of Technology, Cambridge, MA, USA +1
Edge-Efficient Two-Stream Multimodal Architecture for Non-Intrusive Bathroom Fall Detection
Mar 17, 2026Haitian Wang, Yiren Wang, Xinyu Wang +2Fall Detection Multimodal Sensor Fusion
The University of Western Australia, 35 Stirling Highway, Crawley, WA 6009, Australia

Mar 20, 2026J/K move · Enter open · S save

A Multi-Modal CNN-LSTM Framework with Multi-Head Attention and Focal Loss for Real-Time Elderly Fall Detection

Lijie Zhou, Luran Wang

School of Computer Science, University of Nottingham Ningbo China, Ningbo, China · School of Mathematical Sciences, University of Nottingham Ningbo China, Ningbo, China

Fall Detection Wearable Sensing Multimodal Sensor Fusion

Open paper AskPDF arXiv

The increasing global aging population has intensified the demand for reliable health monitoring systems, particularly those capable of detecting critical events such as falls among elderly individuals. Traditional fall detection approaches relying on single-modality acceleration data suffer from high false alarm rates, while conventional machine learning methods require extensive hand-crafted feature engineering. This paper proposes a novel multi-modal deep learning framework, MultiModalFallDetector, designed for real-time elderly fall detection using wearable sensors. Our approach integrates multiple innovations: a multi-scale CNN-based feature extractor capturing motion dynamics at varying temporal resolutions; fusion of tri-axial accelerometer, gyroscope, and four-channel physiological signals; incorporation of a multi-head self-attention mechanism for dynamic temporal weighting; adoption of Focal Loss to mitigate severe class imbalance; introduction of an auxiliary activity classification task for regularization; and implementation of transfer learning from UCI HAR to SisFall dataset. Extensive experiments on the SisFall dataset, which includes real-world simulated fall trials from elderly participants (aged 60-85), demonstrate that our framework achieves an F1-score of 98. 7, Recall of 98. 9, and AUC-ROC of 99. 4, significantly outperforming baseline methods including traditional machine learning and standard deep learning approaches. The model maintains sub- 50ms inference latency on edge devices, confirming its suitability for real-time deployment in geriatric care settings.

Abstract

Healthcare

Explore similar work

A Multi-Modal CNN-LSTM Framework with Multi-Head Attention and Focal Loss for Real-Time Elderly Fall Detection

A Cloud-Edge System for Multimodal Clinical Screening in Resource-Constrained Rural Settings

Edge-Efficient Two-Stream Multimodal Architecture for Non-Intrusive Bathroom Fall Detection