Risk Prediction for human activity by motion inference based on human physical model

2006 Fiscal Year Final Research Report Summary

• Project/Area Number: 17560242
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Single-year Grants
• Section: 一般
• Research Field: Intelligent mechanics/Mechanical systems
• Research Institution: National Institute of Advanced Industrial Science and Technology
• Principal Investigator: MATSUMOTO Osamu National Institute of Advanced Industrial Science and Technology, Intelligent Systems Institute, Senior Research Scientist, 知能システム研究部門, 主任研究員 (10358046)
• Co-Investigator(Kenkyū-buntansha): YAMADA Yoji National Institute of Advanced Industrial Science and Technology, Intelligent Systems Institute, Group Leader, 知能システム研究部門, 研究グループ長 (90166744) ; YOON Seong-Sik National Institute of Advanced Industrial Science and Technology, Intelligent Systems Institute, Post-Doctoral Research Scientist, 知能システム研究部門, 特別研究員 (10443235) ; PARK Un-Sik National Institute of Advanced Industrial Science and Technology, Intelligent Systems Institute, Post-Doctoral Research Scientist, 知能システム研究部門, 特別研究員 (50443213)
• Project Period (FY): 2005 – 2006
• Keywords: Hidden Markov Model(HMM) / Prediction of falling / Motion analysis / Motion Measurement / Welfare Apparatus

Research Abstract

In this research, we aimed to propose a risk prediction method by combining motion prediction method based on human physical model with motion inference method using probability theory and confirm the effectiveness of the proposed method by the experiments using the sensor system attached to human. In order to clarify the problem to be solved, our research target was limited to human falling in case of using a walking aid because we found many reports of the similar accidents. In the first fiscal year (FY 2005), we performed the proposal of the prediction method based on human motion analysis and the risk evaluation method of falling under the physical environmental constraint. Concretely speaking, we constructed human motion measuring system using a magnetic multi-tracking system and proposed the determining method of safety or danger by the data (position, velocity, direction of C.O.G of human) learning system using Hidden Marcov Model (HMM). In the second fiscal year (FY 2006), we performed the investigation of the optimization of sensors' assignment for the prediction of falling, the development of the small, light-weight and wireless sensor system attached to human, and the demonstration of the effectiveness of our proposed risk prediction method. Our developed sensor system includes seven 3D motion sensors with accelerometers, rate gyroscopes and magnetic direction sensor and data processing system, and generates the sound warning to users when determining the high degree of falling risk. As the result, we confirmed the possibility of risk prediction for falling by experiments using the sensor system. The problems to be solved are the reliability of the risk determination, the correspondence to the variations of falling, etc.