Machine learning-based fall detection system for the elderly using passive RFID sensor tags

Abstract

The percentage of elderly people in the world population has been rapidly increasing. Accordingly, the demand for special nursing homes and professional caregivers has also been growing to support the elderly's daily activities. Since elderly people are often unable to get up without assistance after falling, the failure to detect falling accidents can further lead to serious injuries. Hence, early fall detection is crucial to reduce the risk of the elderly's hospitalization and death caused by accidents. In order to promote early fall detection, monitoring services for elderly people based on IoT have been developed. In this paper, the proposed system uses passive RFID sensor tag is composed RFMicron's Magnus S chip, which can measure not only RSSI but also pressure. In our approach, those tags are attached to the indoor footwear and obtain a change of RSSI and pressure values during activity. Our experiment is conducted by extracting features from raw data and classifying activities with machine learning. This paper shows two training models with a different feature set developed in order to evaluate the effectiveness of passive sensor tags. Moreover, the results demonstrate the accuracy of person-dependent and person-independent with the dataset from different subjects.