Integrating Activity Recognition and Nursing Care Records: the System, Experiment, and the Dataset

In this paper, we introduce a system that integrates activity recognition, which collects various activities of the elderly people of a nursing care facility in Japan. This system incorporates various activity levels, sensor data through smartphones, and records by nurses for a period of 4 months. There are 28 activity labels, recorded by a number of staffs/nurses. The system architecture has mobile application to encode data and a networked system in the 6-floored facility, covering a number of residents. The system is designed in such a manner that even a non-expert nurse or user can manage this system. After introducing the entire system and data collection procedure, we recognized the acitivities by exploiting some statistical features and Extremely Randomized Tree. We noticed that there are discrepancies between start and end recording time-stamps. Therefore, we explored a time correction technique. The dataset is an extensive one and it can be explored for various activity recognition in smart-homes or elder care facilities. The dataset includes the sensor data from staffs' smartphones, activity labels, and care details input using the system.