Fall detection in smart home environments using UWB sensors and unsupervised change detection

Abstract

Falls are one of the major issues which can endanger the lives of older adults. Numerous research studies investigate the use of wearable technologies to detect falls in everyday environments. Although wearable sensor solutions provide good accuracy and sensitivity for fall detection, it may not always be convenient or desirable for older adults to wear a tag or sensor in home environments. This paper discusses using non-wearable UWB radar sensors as a practical, environmental fall detection solution in home settings. Specifically, we apply unsupervised change detection methods on UWB sensor data to detect falls. Furthermore, to evaluate the generality of our unsupervised approach, we also apply it to fall detections from accelerometer sensor data. The proposed methods are assessed using the real UWB sensor data sets acquired from the Living Lab at Australian e-Health Research Centre and public available accelerometer sensor data sets. The results show promising outcomes.