Magnetic Induction-based Human Activity Recognition (MI-HAR)

Abstract

Human activity recognition (HAR) is a growing research area with applications in a variety of domains, such as healthcare, rehabilitation, daily life-logging, personal fitness, senior care, and assistance for people with cognitive disorders. Detection of human movements using body sensor networks is one of the most popular yet challenging approaches. There are critical challenges inherent in the sensor-based activity recognition method, such as power management, security, and coverage. Lossy mediums, such as the human body can significantly affect the electromagnetic wave propagating systems. In this paper, we introduce a novel wireless system based on magnetic induction (MI) and deep learning for human activity recognition to address the main challenges and constraints. The generated synthetic MI motion signals are also presented. The proposed MI system is integrated with machine learning techniques to model and detect human activities.