EURO 2024 Copenhagen
Abstract Submission

EURO-Online login

299. Using long short-term memory network for human body depth image motion recognition and abnormal behavior analysis in long-term care environment

Invited abstract in session MA-4: Recent Methodologies in Explainable AI (XAI) 1, stream Recent Advancements in AI .

Monday, 8:30-10:00
Room: 1001 (building: 202)

Authors (first author is the speaker)

1. JongChen Chen
Information Management, National YunLin University of Science and Technology

Abstract

Long-term care has become one of the problems that must be addressed in today's society, and falls are one of the problems often encountered in long-term care. However, most previous studies have used accelerometers to detect falls in older adults. Since this sensor must be placed on the body of the elderly, its effect is quite limited in situations where the willingness to wear it is extremely low. This study uses deep image recognition technology to conveniently track the real-time behavior of the elderly and reduce the problems caused by accidents under the appropriate protection of the subjects' privacy. This research uses a Long Short-Term Memory Network (LSTM) to identify daily human movements and detect abnormal behaviors (such as falls).
In this study, two scenarios were tested. The first is to invite ten healthy people to perform in a laboratory that simulates long-term exposure. Each subject performed nine actions that might occur in the long-term care environment. The second scenario is to apply the database established in the first scenario to an actual long-term photo field. The experimental results show that under the simulated laboratory, the overall performance is quite good. In contrast, in the actual long-term photo field, the performance of motion recognition will be reduced due to the influence of occlusions

Keywords

Status: accepted

Back to the list of papers