Fall Detection Based on the Instantaneous Doppler Frequency: A Machine Learning Approach

Abstract

Modern societies are facing an ageing problem which comes with increased cost of healthcare. A major share of this ever-increasing cost is due to fall related injuries, which urges the development of fall detection systems. In this context, this paper paves the way for building of a radio-frequency-based fall detection system. This paper presents an activity simulator that generates the complex channel gain of indoor channels in the presence of one person performing three different activities, namely, slow fall, fast fall, and walking. We built a machine learning framework for activity recognition based on the complex channel gain. We assess the recognition accuracy of three different classification algorithms: decision tree, artificial neural network (ANN), and cubic support vector machine (SVM). Our analysis reveals that the decision tree, ANN, and cubic SVM achieve an overall recognition accuracy of 73%, 84.1%, and 92.6%, respectively.