RIM: Intensity Restoration of Through-Wall Radar Spectrograms for Human Motion Recognition
IEEE Sensors Journal · 15.05.2026 · 26 (10) · 15808-15820
DOI: 10.1109/jsen.2026.3680096 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Sensors Journal, erschienen am 15.05.2026. Verfasst von Beichen Li und 4 weiteren Autor:innen. Gefördert von: National Natural Science Foundation of China.
Abstract (Originalsprache)
Through-wall radar (TWR) motion recognition is fundamentally limited by intensity distortion, which corrupts the physical intensity relationships that encode human micro-Doppler signatures rather than merely introducing visual artifacts. Although spectrogram restoration can improve recognition, the absence of paired supervision forces existing methods to rely on implicit environmental assumptions, often yielding incorrect intensity mappings that alter motion semantics. In this article, we propose a radar intensity mapping (RIM) framework that explicitly models the intensity relationship between through-wall and free-space spectrograms. RIM learns a geometry-agnostic intensity prior and isolates nonintensity factors to reconstruct physically consistent intensity structures while preserving motionrelevant features. Experiments in two real-world through-wall scenarios show that RIM improves both enhancement quality and recognition accuracy by 4.22% over image restoration baselines and radar-specific enhancement methods. Ablation studies reveal up to 5.94% degradation when either core mechanism is removed. Statistical analysis and visualization evidence further confirm that the reconstructed intensity hierarchy remains physically consistent with freespace behavior, providing direct validation of the proposed intensity modeling strategy for reliable through-wall human motion sensing.
Förderung
- National Natural Science Foundation of China
Zitieren
APA
Li, B., Wang, A., Yang, Y., Dong, T., & He, Y. (2026). RIM: Intensity Restoration of Through-Wall Radar Spectrograms for Human Motion Recognition. IEEE Sensors Journal, 26(10), 15808-15820. https://doi.org/10.1109/jsen.2026.3680096
Vancouver
Li B, Wang A, Yang Y, Dong T, He Y. RIM: Intensity Restoration of Through-Wall Radar Spectrograms for Human Motion Recognition. IEEE Sensors Journal. 2026;26(10):15808-15820. doi:10.1109/jsen.2026.3680096
BibTeX
@article{Li2026,
author = {Li, Beichen and Wang, Anxin and Yang, Yang and Dong, Ting and He, Yuan},
title = {RIM: Intensity Restoration of Through-Wall Radar Spectrograms for Human Motion Recognition},
journal = {IEEE Sensors Journal},
year = {2026},
volume = {26},
number = {10},
pages = {15808--15820},
doi = {10.1109/jsen.2026.3680096}
}
EUVIMED