ARD: accurate and reliable fall detection with using a single wearable inertial sensor

Abstract

Accidental fall is one of the major factors threatening the health of the elderly, which promotes the considerable development of fall detection technology. In our study, a novel method is proposed to detect falls prior to impact during walking. In terms of data collection, acceleration and angular velocity data are collected using a single sensor. By extracting distinctive features, our design goal is to accurately identify fall behavior at an early stage. To improve detection accuracy and reduce false alarms, a classifier based on the joint feature of accelerations and Euler angles (JFAE) analysis is developed. With the support vector machine (SVM) classifier, human activities are classified into eight categories: going upstairs, going downstairs, running, walking, falling forward, falling backward, falling left, and falling right. Not only can it achieve a sensitivity of 96.8% and precision of 96.75%, but also the method we proposed can achieve a high accuracy for the classifier. Compared with the method based on single feature, the method based on multiple features achieves a better performance. The preliminary results indicate that our study has potential application in a fall injury prevention system.