Activity Recognition Based on Kinetic Energy Harvester and Accelerometer
Ling Xiao, Kai Wu · 2018
Harvesting energy from human motion has created an opportunity for battery-free wearable devices to realize long-term human activity monitoring. By simply using the collected motion energy or acceleration is insufficient to distinguish some similar activities very well, it motivates us to exploit the combination of the harvesting kinetic energy and acceleration for activity recognition. An activity recognition approach SRC-EA is proposed, which is based on sparse representation classification to utilize multimodal information. To validate the effectiveness of our approach, a wearable sensor node consists of a commercial energy harvester is developed to collect the harvesting energy and acceleration. We test our approach on a real world data set that covers 10 participants, 9 common activities, and 3 different on-body device positions. The experiment results show that our SRC-EA approach achieves an average accuracy of 92.64% for user-independent activity recognition. Furthermore, our method can not only save energy by reducing sampling rate of accelerometer, but also harvest energy from human motion while the recognition accuracy is maintained at the same level.