Research on the real-time fall prevention system using IMU sensors based on machine learning methods which can be used anywhere on daily basis

2023 Fiscal Year Final Research Report

• Project/Area Number: 21K12798
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 90150:Medical assistive technology-related
• Research Institution: Hokkaido Information University
• Principal Investigator: Toya Nobuyuki 北海道情報大学, 医療情報学部, 教授 (00340654)
• Co-Investigator(Kenkyū-buntansha): 和田 親宗 九州工業大学, 大学院生命体工学研究科, 教授 (50281837) ; 北川 広大 八戸工業高等専門学校, その他部局等, 助教 (20965256)
• Project Period (FY): 2021-04-01 – 2024-03-31
• Keywords: 歩容解析 / 機械学習 / 転倒防止

Outline of Final Research Achievements

In this study, we aimed to realize a simple gait identification system to prevent falls among elderly people, and proposed a method to identify gait by machine learning using wrist movements, which can be easily obtained by wearable sensors, as the main feature. Using walking experiment data, we evaluated the identification performance of the proposed method for “gait with insufficient foot-ground clearance” and “staggering gait” and obtained the accuracy rate higher than 0.9 for all cases. Furthermore, we estimated the “foot clearance value” for each subject using regression analysis, and obtained a correlation coefficient of 0.67 to 0.87 with the actual measured value, confirming the good estimation performance of the proposed method.

Free Research Field

医用生体工学，通信・ネットワーク工学

Academic Significance and Societal Importance of the Research Achievements

本研究の成果によって，手首の加速度データから歩容の判別する提案方式の基本性能と，有効性が確かめられ，この方式を用いた簡易な歩容判別システム実現の可能性が明らかになった．今後提案方式において最適な機械学習アルゴリズムや特徴量を適用することによって，判別や推定の精度をさらに向上させていくことが可能であると考えられる．これにより，屋内外の広い範囲で歩容を確認できる簡易な方式として，高齢者の転倒防止やリハビリテーションへの有効活用が期待できる．