Abstract
This study investigates the possibility of predicting care recipients. living conditions and behaviors using IoT devices and non-contact environmental sensors. We installed these devices into the homes of two care recipients to collect environmental data such as room temperature and illumination levels. We also interviewed them and their caregivers to understand their living situations. Our analysis utilized a computational model incorporating historical weather data and a recurrent neural network (RNN) to forecast environmental conditions and behaviors. The study results showed an accuracy of 90% for waking time, 80% for presence/absence in the room, and 45% for sleeping time. Although there is a need for further improvements in behavior recognition algorithms and cost considerations, this study indicates the potential efficacy of a non-contact monitoring system.
