Design Flow of Neural Network Application for IoT Based Fall Detection System

Abstract

In the remote health monitoring system, it is crucial to identify and analyze the current users' status accurately. The accuracy depends on many different aspects including physical conditions, surrounding environmental conditions, users' distinct features and other factors. In this paper, we investigate the enhacement possibility of IoT based health monitoring system by applying neural network. By training the collected user data from different types of medical emergency-related scenarios, the system would gain better accuracy over the traditional thresholding data analysis systems. In this study, we focus on applying neural network to the fall detection application which involves wireless wearable sensors with accelerometers and a gyroscope. We utilize multilayer perceptron neural network to train user movement datasets including positive falls (falling events) and negative falls (non-falling events). This system design approach has the potential to be extended to multi-purpose user activity and health monitoring system, including people who have potential in needs of medical attentions and daily activity tracking.