Fall Detection for Elderly and Data Classification Movement Activity Using Weighted K-Nearest neighbor Algorithm on a IoT-based Portable Embedded System

Abstract

Detection of falls in the elderly is a continuously developing research topic that requires close monitoring of the health of the elderly. Monitoring the movement activities and detecting the correct fall is, therefore, necessary to avoid the dangers that occur to the elderly. In this paper, we developed a device for detecting falls and classifying the activities of the elderly by using the Internet of Things. The aforementioned allows medical personnel to be able to track the elderlies who are at risk of falling together with predicting the likelihood of falling by the Weight K-Nearest Neighbor method. This system reports the elderly movement activity and detects the falling through the website of IoT cloud provider, including sending crash notification information to healthcare professionals. From testing the fall detection device and tracking the movement activities of the elderly, the device can detect the fall and show the prediction of the occurrence of the fall. The results of the fall detection performance showed that the accuracy of 85.80% and the accuracy of 92.48% in the prediction of the likelihood of falling due to the movement activities. It can classify data of people who are at risk of falling with an average sensitivity of 90.91% and people without the risk of falling with a specific average of 98.98%. In case of an emergency, the medical personnel can use this information as a reference in association with the proper diagnosis of falls in the elderly.