Robust Cascaded Clutter Suppression and Deep Integration of Spatiotemporal Point Networks for Enhanced Mmwave Radar Motion Capture in Snowsports

Abstract

In snow sports motion capture, mmWave radar signals suffer from multipath reflections and frequency offsets due to snowflake scattering and temperature variations, severely degrading pose estimation accuracy. To address this, we propose a cascaded anti-interference framework composed of adaptive MTI filtering, genetic sparse array optimization, and hybrid carrier tracking. These physical-layer enhancements are followed by a spatiotemporal 3D CNN–LSTM network for motion decoding and a multimodal Kalman-particle filter for trajectory fusion. Experimental validation in both simulation and real-world snow environments confirms the framework’s robustness. Compared to baseline systems, the proposed method reduces the joint positioning root mean square error (RMSE) by up to 72%, enhances angular velocity tracking precision by 72%, and improves signal-to-noise ratio (SNR) by 24.3 dB. The end-to-end processing delay remains under 26 ms, ensuring real-time deployment. These results demonstrate significant improvements in accuracy, robustness, and real-time performance under harsh environmental interference, offering a viable solution for mmWave-based motion capture in snowy sports scenarios.