Low Power Human Behavior Detection Method with Low Sampling Frequency 3D-Accelerometer

Recently, health is paid attention more and more, exercise has been become a part of daily life. Meanwhile, with the development of sensors and mechanical intelligence, the sensor began to be is a popular approach to help people daily work, so the importance of research is to use the sensor within a simple method and low power. In this paper, the LPHBD - Low Power Human Behavior Detection Method with Low Sampling Frequency 3D-Accelerometer has been presented. In this system, only one low sampling rate accelerometer (only 5.5Hz) was used to detect the human behaviors like running, walking, sitting and so on, the low sampling rate is in order to save energy and simplify the algorithm, at the same time, maximize the use of machine learning so that it has increased the breadth and applicability, when it cannot too much waste the equipment resource. The experiment results show that behavior detection accuracy is almost greater than 95%, except accuracy of running behavior detection, which is about 86.4%.