Channel Modeling and Characteristics of Human Motion Using Dynamic Ray Tracing and MoCap
Yuan Liu, Johann Fuchs, Mallikarjun Shankar · 2024
Radio-based motion behavior recognition is a key to enabling integrated sensing and communication (ISAC) in the 6G landscape. Data-driven approaches are popular to achieve human motion identification and most of them are based on the range-Doppler maps (RDMs) of radar measurements, where the Doppler and the micro-Doppler (uDop) frequency shifts due to motion are embedded. In this paper, we characterize the human motions of the radar signals using typical wireless channel modeling parameters, e.g., normalized Doppler power spectral density (ND-PSD) and root-mean-square Doppler spread (RMS- DS). Specifically, a proposed dynamic ray tracing (RT) based on Blender is used to simulate time-varying radar channels, where the real human motions are imported from the motion capture (MoCap) database, hence the simulated radar channels contain dynamic real human motion, multipaths of clutters, and also the interaction between target and the environment. The simulation results show that the Doppler channel parameters are consistent with the conventional radar images, besides containing more semantic information about the human motions and scenarios. This provides the feasibility of utilizing channel parameters as training datasets.