3-D Signature Reconstruction and Activity Classification Using Distributed 60-GHz MIMO Radar
IEEE Transactions on Radar Systems · 2026 · 4 · 1291-1304
DOI: 10.1109/trs.2026.3712784 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Radar Systems, 2026. Verfasst von Kuitong Lou und 2 weiteren Autor:innen. Die Arbeit nennt 39 Literaturangaben. Gefördert von: Dutch Research Council (NWO) with their Open Technology Programme (OTP) grant Distributed Adaptive Radar for Enhanced Sensing and Classification.
Abstract (Originalsprache)
Distributed radars offer the potential to overcome the limited field of view and ambiguities of single radar systems. However, effectively fusing data from multiple radar nodes remains challenging, particularly for reliable 3-D signature reconstruction and classification of human activities under diverse motion conditions. To address these challenges, we propose a distributed fusion method for 3-D signature reconstruction and human activity recognition (HAR) using distributed 60-GHz multiple-input multiple-output (MIMO) radar nodes. The proposed approach incorporates an adaptive radar signal processing pipeline designed to tackle both static and dynamic movements. At each radar node, Capon beamforming is applied to generate radar cubes in range, azimuth, and elevation, which are then fused to reconstruct 3-D intensity maps and voxelwise velocity vectors of the observed extended human target. The reconstructed signatures, consisting of 3-D intensity and Doppler vectors, are then used as inputs to deep learning networks for HAR. A hybrid CNN-bidirectional long short-term memory (BiLSTM) model is developed to capture both spatial and temporal features and realize the classification of activities in a sequence-by-sequence fashion. The method was evaluated with a dataset collected using a network of three 60-GHz radars and including seven activities performed by 30 participants. The results show that both static postures and dynamic activities can be captured effectively by the reconstructed 3-D signatures. The proposed method can also handle the diversity of movements and achieves an average macro- $F1$ score of 90.9% for a four-class task under a leave-one-person-out (L1PO) evaluation, and 92.0% under subjectwise cross-validation for the same four-class task. Furthermore, ablation studies show that comparable HAR performance is achieved using only two out of the three radars, provided that sufficient spatial diversity is ensured. This suggests that the viewing and deployment geometry of the radar nodes is more critical than their number in absolute terms.
Förderung
- Dutch Research Council (NWO) with their Open Technology Programme (OTP) grant Distributed Adaptive Radar for Enhanced Sensing and Classification
Zitieren
APA
Lou, K., Yarovoy, A. G., & Fioranelli, F. (2026). 3-D Signature Reconstruction and Activity Classification Using Distributed 60-GHz MIMO Radar. IEEE Transactions on Radar Systems, 4, 1291-1304. https://doi.org/10.1109/trs.2026.3712784
Vancouver
Lou K, Yarovoy AG, Fioranelli F. 3-D Signature Reconstruction and Activity Classification Using Distributed 60-GHz MIMO Radar. IEEE Transactions on Radar Systems. 2026;4:1291-1304. doi:10.1109/trs.2026.3712784
BibTeX
@article{Lou2026,
author = {Lou, Kuitong and Yarovoy, Alexander G. and Fioranelli, Francesco},
title = {3-D Signature Reconstruction and Activity Classification Using Distributed 60-GHz MIMO Radar},
journal = {IEEE Transactions on Radar Systems},
year = {2026},
volume = {4},
pages = {1291--1304},
doi = {10.1109/trs.2026.3712784}
}
EUVIMED