Metraipolis › archive › dc_4e4e5e9bebd3

Bidirectional Modeling of mmWave Human Signals for Enhancement & Synthesis — arXiv:2611.08870

Source: arXiv (en)
Original: https://arxiv.org/abs/2611.08870
License (as archived): CC0-1.0
Archived: 2026-11-09T22:40:54.822Z
SHA-256 of the archived text: 17969ce5b67bf3d32ad6efdbfe8794db1ea215e523971881074bd9e1446e903e
Stable URL of this page: https://metraipolis.com/documento/dc_4e4e5e9bebd3

How to cite

Bidirectional Modeling of mmWave Human Signals for Enhancement & Synthesis — arXiv:2611.08870 — arXiv (en), https://arxiv.org/abs/2611.08870 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_4e4e5e9bebd3 (Metraipolis archive dc_4e4e5e9bebd3, sha256 17969ce5b67bf3d32ad6efdbfe8794db1ea215e523971881074bd9e1446e903e)

BibTeX

@misc{dc4e4e5e9bebd3,
  title = {Bidirectional Modeling of mmWave Human Signals for Enhancement & Synthesis — arXiv:2611.08870},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.08870},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_4e4e5e9bebd3; sha256 17969ce5b67bf3d32ad6efdbfe8794db1ea215e523971881074bd9e1446e903e},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.08870. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Millimeter-wave (mmWave) is attractive for human-centered robotics in smoke, darkness, and privacy-sensitive indoor spaces: it measures reflections rather than visually identifiable imagery and remains usable without light. Deployed systems, however, face two complementary needs that run in opposite directions. Enhancement starts from a raw range-angle-Doppler (RAD) heatmap, where human returns coexist with nonhuman structure, and aims to isolate the human signal by removing nonhuman returns, yielding a clean RAD input for downstream perception. Synthesis starts from denser human geometry and aims to synthesize natural RAD observations, including environmental and hardware features, thereby expanding scarce supervision data. Although the two tasks reverse direction, both require modeling how human and nonhuman reflections coexist in the same measurement. Among RAD axes, Doppler provides motion information native to the modality, with the potential to support this modeling. We therefore cast both tasks on the dual-domain RAD heatmap and propose HeatWave, which injects Doppler when it is locally reliable, and routes Doppler asymmetrically for enhancement and synthesis within one shared backbone. This design improves both tasks: enhanced observations reduce mean per-joint position error by 11.4% relative to the raw recording, and synthesized samples that replace half of the real radar labels recover most of the activity accuracy lost to the reduced budget. Relative to an unspecialized dual-task heatmap baseline, the enhancement improves motion-critical samples under metallic clutter by 2.3times more than when geometry already suffices, and synthesis follows the same ordering with a narrower spread, providing the community with a research direction more closely aligned with the signal's characteristics.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.HC, nuovi lavori del giorno).
Categorie arXiv: Human-Computer Interaction (cs.HC).
Pagina del lavoro: https://arxiv.org/abs/2611.08870 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-11-09T22:33:32.075Z dalla pagina https://arxiv.org/list/cs.HC/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).