Context-aware emergency detection method for edge computing-based healthcare monitoring system

When taking physical exercise, healthcare monitoring systems are needed to detect emergencies, such as sychnosphygmia, and so on. However, heterogeneity and privacy make it more challenging to collect and process multi-source healthcare data in a real time manner. To support healthcare monitoring, in this research, a multi-layer edge computing-based framework is designed to detect emergencies in low latency and secure ways. First, a universal data model is defined to process heterogeneous data from multi-sources such as sensors or smart devices. Then, Gated Recurrent Units-based (GRU) model is utilized to capture and process physiological data series. Fine-grained detection models are built and distinguished according to scenario information and individual health persona. In the proposed edge computing-based framework, detection algorithm computation and physiological index data storage are close to end user sides to support low latency. Meanwhile encryption algorithm is used to protect data privacy. Finally, a case study in the Physical Education Class scenario is implemented to demonstrate the feasibility of our methods. The experimental result shows that our GRU model is more accurate and is three times faster for emergency identification compared with Support Vector Classification model.