The Trade-off between Energy-Accuracy in the IoT-based Activity Monitoring System

Abstract

Activity monitoring system is used in many fields such patient’s activity monitoring system for self-quarantine in their home. The IoT- based activity monitoring system uses the limited resources (e.g., bandwidth, battery and memory) for monitoring the user’s activity. The limited resources (such as battery) provide the limited lifetime battery in activity monitoring system. By resource efficiency, it will extend the battery lifetime. Resource efficiency is achieved by adaptively reporting user activity depending on the level of the user’s activity emergency. But, when the user’s activity reporting data is based on the emergency level, then it reduces the data detail and its activity recognition accuracy. So, we develop energy-savings techniques for user’s activity reporting and analyze the effect of energy-savings techniques to the accuracy of activity recognition using different methods. The results show the energy-savings techniques can save battery life up to 8%, bandwidth up to 146,5 bytes/sec and memory up to 2,8% compared to non-energy saving technique. But the energy-saving techniques give less accuracy in the four different activity recognition methods up to 11% in average.