Human Activity Recognition Using Machine Learning Based Mobile Smart Phone Magnetometers

摘要

As the performance of sensors embedded in mobile smart phones has improved, many studies using data collected from sensors are being conducted. In this study, using the data obtained from the 3-axis magnetic sensor mounted on the smartphone, a study on the recognition of four human activities was performed using machine learning. From the total data of the 3-axis magnetic sensor, the data was bundled into frames for 2 seconds, divided into several frames, and then supervised learning was carried out using it as an input to the convolutional neural network. The operation of the magnetic sensor depending on the direction was confirmed, and the human activity recognition for standing, sitting, walking, and jogging was verified.