Deep Learning of Biomechanical Dynamics in Mobile Daily Activity and Fall Risk Monitoring*

Smart health is paving a promising way for modern health management. Daily activity and fall risk monitoring is one important application that urges smart technologies, resulting from the fact that there are 29 million falls and 7 million fall injuries per year, and also the fact that appropriate exercise can lower the risk of death by up to 20 to 70%. However, it is very challenging to accurately identify an activity due to the diversity of the human biomechanical dynamics. Main reasons include: even a same person usually has different motion characteristics when performing a same activity; there are many different activities in our daily lives; and the sensor wearing habit may be different. In this paper, focusing on these challenges, a new intelligent computational approach is proposed for robust activity detection, leveraging biomechanical dynamics enhancement and deep learning technologies. It can unveil deep hidden biomechanical patterns from the mobile phone-sensed motion data, and robustly detect 17 types of daily and fall activities performed by 30 people. The detection accuracy of 11,770 activities is as high as 93.9%, indicating the effectiveness of the proposed approach. This research is expected to greatly advance mobile daily activity and fall risk monitoring in smart health era.

[1]  Giuseppe De Pietro,et al.  A smart mobile, self-configuring, context-aware architecture for personal health monitoring , 2018, Eng. Appl. Artif. Intell..

[2]  Shuang Zhu,et al.  Real-time Activity and Fall Risk Detection for Aging Population Using Deep Learning , 2018, 2018 9th IEEE Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON).

[3]  Fan Yang,et al.  Real-time human activity classification by accelerometer embedded wearable devices , 2017, 2017 4th International Conference on Systems and Informatics (ICSAI).

[4]  A. Thilmann,et al.  The limitations of the tendon jerk as a marker of pathological stretch reflex activity in human spasticity. , 1993, Journal of neurology, neurosurgery, and psychiatry.

[5]  Wan Young Chung,et al.  Activity monitoring from real-time triaxial accelerometer data using sensor network , 2007, 2007 International Conference on Control, Automation and Systems.

[6]  Youngbum Lee,et al.  Implementation of Accelerometer Sensor Module and Fall Detection Monitoring System based on Wireless Sensor Network , 2007, 2007 29th Annual International Conference of the IEEE Engineering in Medicine and Biology Society.

[7]  Tim Lüth,et al.  Accelerometer based real-time activity analysis on a microcontroller , 2011, 2011 5th International Conference on Pervasive Computing Technologies for Healthcare (PervasiveHealth) and Workshops.

[8]  Dian Zhou,et al.  Deep Arm/Ear-ECG Image Learning for Highly Wearable Biometric Human Identification , 2017, Annals of Biomedical Engineering.

[9]  Daniela Micucci,et al.  UniMiB SHAR: a new dataset for human activity recognition using acceleration data from smartphones , 2016, ArXiv.