Search papers, labs, and topics across Lattice.
This paper introduces a physics-guided preprocessing framework for millimeter-wave (mmWave) human pose estimation (HPE) that leverages explicit physical priors to improve efficiency and accuracy. The framework models inter-dimensional correlations in mmWave data by coupling range and angle, leveraging Doppler for motion continuity, and applying multi-scale fusion aligned with the human body. Experiments demonstrate that this approach reduces the number of parameters by 55.7-88.9% compared to existing mmWave baselines while maintaining competitive accuracy and enabling real-time deployment on a Raspberry Pi.
By explicitly encoding mmWave sensing physics, a new HPE preprocessing method slashes parameter counts by up to 89% while matching accuracy, proving that sometimes the best AI is less AI.
We revisit millimeter-wave (mmWave) human pose estimation (HPE) from a signal preprocessing perspective. A single mmWave frame provides structured dimensions that map directly to human geometry and motion: range, angle, and Doppler, offering pose-aligned cues that are not explicitly present in RGB images. However, recent mmWave-based HPE systems require more parameters and compute resources yet yield lower estimation accuracy than vision baselines. We attribute this to preprocessing modules: most systems rely on data-driven modules to estimate phenomena that are already well-defined by mmWave sensing physics, whereas human pose could be captured more efficiently with explicit physical priors. To this end, we introduce processing modules that explicitly model mmWave's inter-dimensional correlations and human kinematics. Our design (1) couples range and angle to preserve spatial human structure, (2) leverages Doppler to retain human motion continuity, and (3) applies multi-scale fusion aligned with the human body. A lightweight MLP is involved as the regressor. In experiments, this framework reduces the number of parameters by 55.7-88.9% on the HPE task relative to existing mmWave baselines while maintaining competitive accuracy. Meanwhile, its lightweight nature enables real-time Raspberry Pi deployment. Code and deployment artifacts will be released upon acceptance.
