EUVIMEDEUVIMED entdecken
Crossref

End-to-End Through-Wall Respiratory Waveform Reconstruction for 60-GHz MIMO FMCW Radar

Abdel-Kareem Moadi, Chandler Bauder, Vijaysrinivas Rajagopal, Aly E. Fathy

IEEE Journal of Electromagnetics, RF and Microwaves in Medicine and Biology · 2026-09 · 10 (3) · 375-385

DOI: 10.1109/jerm.2026.3683327 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Journal of Electromagnetics, RF and Microwaves in Medicine and Biology, 2026-09. Verfasst von Abdel-Kareem Moadi und 3 weiteren Autor:innen. Die Arbeit nennt 27 Literaturangaben. Gefördert von: National Science Foundation Industry-University Collaborative Research Center.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This paper presents an end-to-end deep-learning framework that reconstructs respiratory waveforms for a seated subject in through-wall scenarios using 60 GHz MIMO FMCW radar. Conventional through-wall vital-sign pipelines that include localization, spatial filtering, phase extraction, and downstream biomarker analysis, are fragile under wall-induced attenuation and dispersion. This usually promotes lowering the operating frequency yielding inaccurate localizations and noisy waveforms. To address these limitations while retaining the spatial resolution of mmWave radar, we develop a single deep neural network that operates directly on multi-channel range-domain I/Q data and outputs the corresponding time-domain respiration waveform, implicitly performing localization, spatial filtering, and temporal reconstruction. The model is trained and evaluated on 720 minutes of data from three separate subjects positioned at ranges of 1–2 m and azimuths within $\pm \mathbf{60}^{\circ }$ behind drywall, cement board, glass, and wood panels of varying thicknesses. Across all wall types, ranges, and angles, the network attains a mean cosine similarity of 0.92 between reconstructed and ground-truth waveforms and maintains 0.91 on a held-out glass wall material, indicating robustness to moderate changes in wall composition within this experimental setting. Relative to baseline methods, the proposed approach reduces waveform mean-squared error by at least 43% and decreases errors in derived biomarkers, breathing rate, inter-breath interval, and inhalation/exhalation ratio, by up to 69%. These results demonstrate that end-to-end learning on 60 GHz MIMO FMCW radar can provide accurate through-wall respiration waveforms suitable for downstream biomarker analysis while avoiding hand-engineered localization and phase-extraction stages.

Förderung

  • National Science Foundation Industry-University Collaborative Research Center

Zitieren

APA

Moadi, A., Bauder, C., Rajagopal, V., & Fathy, A. E. (2026). End-to-End Through-Wall Respiratory Waveform Reconstruction for 60-GHz MIMO FMCW Radar. IEEE Journal of Electromagnetics, RF and Microwaves in Medicine and Biology, 10(3), 375-385. https://doi.org/10.1109/jerm.2026.3683327

Vancouver

Moadi A, Bauder C, Rajagopal V, Fathy AE. End-to-End Through-Wall Respiratory Waveform Reconstruction for 60-GHz MIMO FMCW Radar. IEEE Journal of Electromagnetics, RF and Microwaves in Medicine and Biology. 2026;10(3):375-385. doi:10.1109/jerm.2026.3683327

BibTeX
@article{Moadi2026,
  author = {Moadi, Abdel-Kareem and Bauder, Chandler and Rajagopal, Vijaysrinivas and Fathy, Aly E.},
  title = {End-to-End Through-Wall Respiratory Waveform Reconstruction for 60-GHz MIMO FMCW Radar},
  journal = {IEEE Journal of Electromagnetics, RF and Microwaves in Medicine and Biology},
  year = {2026},
  volume = {10},
  number = {3},
  pages = {375--385},
  doi = {10.1109/jerm.2026.3683327}
}
Kostenlos starten →