Internet of things based multi-sensor patient fall detection system

Abstract

Accidental falls of patients cannot be completely prevented. However, timely fall detection can help prevent further complications such as blood loss and unconsciousness. In this study, the authors present a cost-effective integrated system designed to remotely detect patient falls in hospitals in addition to classifying non-fall motions into activities of daily living. The proposed system is a wearable device that consists of a camera, gyroscope, and accelerometer that is interfaced with a credit card-sized single board microcomputer. The information received from the camera is used in a visual-based classifier and the sensor data is analysed using the k-Nearest Neighbour and Naïve Bayes’ classifiers. Once a fall is detected, an attendant at the hospital is informed. Experimental results showed that the accuracy of the device in classifying fall versus non-fall activity is 95%. Other requirements and specifications are discussed in greater detail.