Device-Free Activity Recognition Using Ultra-Wideband Radios

Human Activity Recognition (HAR) is a fundamental building block for the current trend of smart devices in Internet of Things (IoT). Ultra-Wideband RF technology has been used in localization research while Wi-Fi Channel State Information (CSI) has been widely investigated for non-obtrusive activity recognition in the literature. This paper investigates the feasibility of using UWB technology for Human Activity Recognition (HAR). The key idea is to use machine learning classification algorithms most suited to train models to classify different activities using the Channel Impulse Response (CIR) data of the UWB signals. Our experiments show that by using CIR data as features we can classify simple activities such as standing, sitting, lying with an accuracy of 95%. To compare this performance, we have also trained statistical models using Wi-Fi CSI. We found that, for all models UWB CIR significantly outperformed Wi-Fi CSI. Thus, we believe UWB to be a very effective technology in the context of device-free activity recognition.