Study on motion recognition scheme using acceleration sensor in mobile devices

Abstract

In this paper, a motion recognition scheme using acceleration sensor of smart devices is proposed and is experimentally analyzed with a server. Sensor data of 3-axis acceleration of a smart device is collected and 2nd order Butterworth low pass filter (LPF) is applied to reduce noise. With the collected sensor data, the various motions such as falling, sitting, lying and walking are distinguished one another. Based on the fixed threshold by the probability approach, falling motion is especially distinguished from other motions as an indication of emergency situation. A server displays current states and alarm states. The proposed scheme is evaluated through experiments. The results of experiments show that the accuracy of motion recognition is more than 94% in all the motions. As for a falling recognition, the accuracy is 96% ©2013 SERSC.