Abstract:
With people's busy schedules, elderly people have to
stay alone in their houses in the daytime. There are a large
number of accidents have happened to elderly people when they
are alone at home. It is crucial to have a monitoring system to
identify the potential hazards for the protection of elders to
address this risk. In this research, we propose a method to
identify postural behaviors, walking abnormalities, and falling
situations using the skeleton data obtained from the Microsoft
Kinect camera. In this paper, we discuss the identification of the
falling of an older person. For that, we used an LSTM model,
and the features of the model are velocities of angles and joints
of the skeleton. This system achieved a validation accuracy of
88.34%, and it offers a promising solution for keeping an eye on
and recognizing potential dangers for elderly people.
Citation:
S. A. T. N. Sudasinghe, I. K. S. Sooriyabandara, A. H. M. D. P. M. Banadara, H. Rajendran and A. G. B. P. Jayasekara, "Vision Attentive Robot for Elderly Room," 2023 Moratuwa Engineering Research Conference (MERCon), Moratuwa, Sri Lanka, 2023, pp. 19-24, doi: 10.1109/MERCon60487.2023.10355403.