What Physics do Data-Driven MoCap-to-Radar Models Learn?

数据驱动的动作捕捉转雷达模型究竟学到了什么物理规律？

Abstract: Data-driven MoCap-to-radar models generate plausible micro-Doppler spectrograms, but do they actually learn the underlying physics? We introduce a physics-based interpretability framework to answer this question via two proposed complementary metrics: one measures alignment between model predictions and the physics-derived Doppler frequency, while the other tests whether predictions preserve the velocity-frequency relationship under velocity intervention. Both metrics require only MoCap input and model predictions, without access to measured radar data.

摘要： 数据驱动的动作捕捉（MoCap）转雷达模型能够生成看似合理的微多普勒频谱图，但它们是否真的学到了底层的物理规律？为了回答这个问题，我们引入了一个基于物理的可解释性框架，并通过两个互补的指标进行评估：第一个指标衡量模型预测与物理推导出的多普勒频率之间的一致性；第二个指标则测试在速度干预下，预测结果是否能保持速度与频率之间的关系。这两个指标仅需动作捕捉输入和模型预测结果，无需访问实际测量的雷达数据。

Experiments across several model architectures reveal that low reconstruction error does not guarantee physical consistency: some, but not all, models achieve low error yet perform poorly on the two physics-based metrics. Further analysis shows that temporal attention is critical for transformer-based models to learn the underlying physics.

针对多种模型架构的实验表明，较低的重构误差并不能保证物理一致性：部分模型虽然实现了低误差，但在上述两个基于物理的指标上表现不佳。进一步分析显示，对于基于 Transformer 的模型而言，时间注意力机制（temporal attention）对于学习底层物理规律至关重要。

---

Paper Details:

• Authors: Kevin Chen, Kenneth W. Parker, Anish Arora
• arXiv ID: 2605.00018
• Subjects: Machine Learning (cs.LG); Signal Processing (eess.SP)
• Submission Date: 19 Apr 2026

论文详情：

• 作者： Kevin Chen, Kenneth W. Parker, Anish Arora
• arXiv ID： 2605.00018
• 学科分类： 机器学习 (cs.LG)；信号处理 (eess.SP)
• 提交日期： 2026年4月19日