Device-Free Kick Detection Using Ultra-Wideband Channel Impulse Response

Ultra-Wideband (UWB) RF technology has been widely used in localization research, but its potential for human activity detection has been largely overlooked. In this paper, we propose a novel approach for kick detection using UWB channel impulse response (CIR). Our approach involves utilizing both the magnitude and phase of CIR data as inputs to a neural network, and stacking the data in time to form a 2-dimensional input, similar to an image. Experimental results show that our method achieves a classification accuracy of 96.7% for kick and non-kick detection in a laboratory environment, and 70% true positive and 5% false positive in real car deployment using existing UWB key infrastructure with no additional hardware cost. This study demonstrates the potential of UWB technology for human activity detection beyond its traditional use in localization research.