A wearable sensor-based activity prediction system to facilitate edge computing in smart healthcare system

An increase in world population along with elderly people is causing fast rises in healthcare costs. Technologies (e.g., Internet-of-Things, Edge-of-Things, and Cloud-of-Things) in healthcare systems are going through a transformation where health monitoring of people is possible without hospitalization. The advancement of sensing technologies helps to make it possible to develop smart systems to monitor human behaviors continuously. In this work, a wearable sensor-based system is proposed for activity prediction using Recurrent Neural Network (RNN) on an edge device (i.e., personal computer or laptop). The input data of the system are obtained from multiple wearable healthcare sensors such as electrocardiography (ECG), magnetometer, accelerometer and gyroscope sensors. Then, an RNN is trained based on the features. The trained RNN is used for predicting the activities. The system has been compared against the conventional approaches on a publicly available standard dataset. The experimental results show that the proposed approach outperforms other traditional methods. Graphics Processing Unit (GPU) in the edge device is utilized to take the advantage of fast computation of experimental data.