Inexpensive cameras and face recognition technology are paving the way towards capturing an individual’s behavior without the use of wearable devices, eliminating the inconvenience of users having to wear sensors and routinely exchange/recharge batteries. Precision care at nursing homes, such as fall prevention and management of medicine, requires individual monitoring of walking status in homes with multiple elderly residents. This study demonstrates a new system for daily walk monitoring of individual elderly over the long term without carrying devices or exchanging batteries.
We developed a system to individually monitor the walking speed of elderly people. The system consists of RGB-D cameras (Microsoft’s Kinect), a face recognition engine, a registered person identification engine with deep learning technology (Res-Net), and individualized data storage. The developed system allowed us to collect walking speed, posture data, RGB-D image data, and important events such as falls for each person. The authors installed the developed system in one nursing home to evaluate its effectiveness in an actual nursing home environment. The test home had eight elderly adults (79 to 104 years old) and seven nursing staff.
The system succeeded in distinguishing the target person from others and in monitoring the walking speed of each elderly person. Using the system, we could analyze changes in walking speed by day and week. Walking speed ranged from 0.2 to 1.2 m/s. The error rate of person identification was within 2.9% and the error rate of walking speed was within 4.9%.
The experimental results indicate that the developed system is useful for monitoring individual elderly in an actual nursing home in terms of both ease of operation and accuracy of data.