EdgeCNN: A Hybrid Architecture for Agile Learning of Healthcare Data from IoT Devices

Jian Yu,Bin Fu,Ao Cao,Zhenqian He,Di Wu

DOI: https://doi.org/10.1109/PADSW.2018.8644604

2018-01-01

Abstract:With the pervasive usage of IoT devices to collect healthcare data from human body, the real-time need on data acquirement and analysis for effective diagnosis and feedbacks becomes an challenging issue in practice. We propose a hybrid architecture, called EdgeCNN, that balances the capability of edge and cloud computing to address this issue for agile learning of healthcare data from IoT devices. Specifically, deep learning is customized as the inference method running on edge devices, making real-time analysis and diagnosis closer to the IoT data source. This can significantly reduce learning latency and network I/O, ease the pressure on the cloud platform for large user groups and massive data, and drastically decrease the cost to build and maintain cloud platforms. To verify the feasibility of EdgeCNN, we design a set of streamlined diagnosis model and learning algorithm for edge computing based on convolutional neural network (CNN), facilitating EdgeCNN to identify and infer electrocardiograms in real-time, as a specific healthcare application using smart devices on the edge. Our experimental results show that under the premise of ensuring the accuracy, EdgeCNN has significant advantages in diagnosis delay, network I/O, application usability and resource cost, in comparison with the architecture solely based on the cloud computing. Another important benefit from EdgeCNN is that it can effectively protect the privacy of user data from IoT devices.