Development of a monitoring system of preventing falls from a bed using deep learning to recognize behavior for elderly people

2019 Fiscal Year Final Research Report

• Project/Area Number: 16K20847
• Research Category: Grant-in-Aid for Young Scientists (B)
• Allocation Type: Multi-year Fund
• Research Field: Gerontological nursing
• Research Institution: National Institute of Information and Communications Technology (2017-2019); Kochi National College of Technology (2016)
• Principal Investigator: HIRONOBU SATOH 国立研究開発法人情報通信研究機構, ナショナルサイバートレーニングセンターサイバートレーニング研究室, 主任研究員 (90461384)
• Project Period (FY): 2016-04-01 – 2020-03-31
• Keywords: 深層学習 / 行動認識 / 高齢者介護 / 深度センサ / HCI

Outline of Final Research Achievements

Elderly people are more likely to suffer serious injuries such as broken bones and become bedridden if they fall from a bed. Therefore, the proposed system can recognize the signs of a fall from the continuous behavior of the elderly person on the bed, such as "flapping his legs" or "sitting on the bed" by an artificial intelligence. The proposed system of this research uses a Kinect to sense the body of person on the bed, and the DBN, which uses the data sensed by the Kinect as input information, understands the behavior and anticipate the fall. The behavior leading to falls is reported by health care professionals to be characteristic for each subject. Therefore, it is necessary to use data containing characteristic behaviors for learning. The realization of this mechanism allowed for the realization of individualized predictors of a fall.

Free Research Field

パターン認識

Academic Significance and Societal Importance of the Research Achievements

医療機関における身体拘束は難しく，介護者による24時間の見守り介護が必要とされた。しかし，我が国の医療現場では，介護士不足が深刻であり， 24時間の見守り介護は難しい。介護士に代わり、人工知能によりモニタリングすることが本研究の狙いである。このシステムの実現により、事故防止によるQOLの向上、介護問題の解決およびわが国の医療費の抑制に繋がる。また、提案研究は，ヒトの行動を理解する知能マシンの創出と，人工知能研究において重要な意味を持つ，数少ない研究成果である。