A Radar-Based Concept for Simultaneous High-Resolution Imaging and Pixel-Wise Velocity Analysis for Tracking Human Motion

Abstract

The radar-based analysis of human motion is actively being researched due to its contact- and markerless nature and ability to measure motion directly via the Doppler effect. Especially in medical and biomechanical fields, precise movement analysis is crucial. However, existing radar-based studies typically exhibit low lateral resolution, focusing on velocity evaluations and the tracking of scattering centers resolvable in the range or Doppler domains. In this work, we present a novel concept that enables a pixel-wise velocity analysis of human motion in radar near-field imaging scenarios. For this, we utilize the well-established back-projection technique to reconstruct consecutive radar images and perform a subsequent pixel-wise phase comparison. To accurately capture pixel-specific velocities along the depth dimension, this is followed by corrections of near-field geometry distortions accounting for aperture properties and pixel positions. Our theoretical derivations are supported by comprehensive point target simulations. To assess the performance of the proposed approach, we conducted a proof-of-concept study. We tracked a hand surface's movement while performing a finger tapping motion and compared the fingertip position and velocity determined by the radar with the respective values obtained from an optical marker-based system. The results showed a velocity measurement accuracy of $8.1 \,\mathrm{mms}^{-1}$ and a tracking accuracy of $1.4 \,\mathrm{m}\mathrm{m}$ , demonstrating the great potential of our approach. The high angular resolution of the velocity measurement enables the tracking of the entire illuminated body shell, extending the range of future applications of radar-based motion analysis.