Real-time human activity classification by accelerometer embedded wearable devices

Abstract

Human activity information can be used in a lot of applications such as fitness monitoring. In this paper an activity classification wearable device system is designed which can provide the activity information of the users via a three-axis kinematic sensor. The features in time domain and frequency domain of acceleration data are extracted, and Decision Tree algorithm is applied. The training module is performed once off line to generate the classification model, while the classification can be executed real time on the STM32L low-power microcontroller. The wearable device was worn in watch style in experiment and it offered activity information in acceptable accuracy.