Affiliations: [a] School of Automation, Xi’an University of Posts and Telecommunications, Xi’an, China
| [b] Communication and Information Engineering, Xi’an University of Posts and Telecommunications, Xi’an, China
| [c] School of Informatics, Computing and Cyber Systems, Northern Arizona University, Flagstaff, USA
Correspondence:
[*]
Corresponding author. Junsuo Qu, School of Automation, Xi’an University of Posts and Telecommunications, Xi’an, China. E-mail: [email protected].
Abstract: In this paper, we designed a smart bracelet to realize human behavior recognition, remote monitoring, fall detection, alarm and other functions. We use a six-axis inertial measurement module to collect a large amount of human body posture data. The Kalman filter is used to preprocess the data of acceleration and gyroscope, and then the processed series data is segmented by sliding window method. At last these data are put into a convolutional neural network(CNN) to identify human activities. Furthermore, we designed a two-level fall behavior judgment, the first-level threshold determination is performed through smart bracelet, when a suspected fall behavior occurred, the sequence data are uploaded to the cloud platform, the second CNN discriminant model is triggered to accurately determine whether a fall behavior occurs.
Keywords: Convolutional neural network, human behavior recognition, fall detection
