Radar fall motion detection using deep learning

Abstract

Radar has a great potential to be one of the leading technologies to perform in-home monitoring of elderly. Radar signal returns corresponding to human gross-motor activities are nonstationary in nature. As such, time-frequency (TF) analysis plays a fundamental role in revealing constant and higher order velocity components of various parts of the human body under motion which are important for motion discrimination. In this paper, we consider radar for fall detection using TF-based deep learning approach. The proposed approach learns and captures the intricate properties of the TF signatures without human intervention and feeds the underlying features to the classifier. Experimental data is used to demonstrate the effectiveness of the proposed fall detection deep learning approach in comparison with the principal component analysis method and techniques incorporating manual selections of a few dominant features.