Robust Low-Snapshot DOA Estimation for Sparse Arrays via a Hybrid Convolutional Graph Neural Network

We propose a hybrid Convolutional Graph Neural Network (C-GNN) for direction-of-arrival (DOA) estimation in sparse sensor arrays under low-snapshot conditions. The C-GNN architecture combines 1D convolutional layers for local spatial feature extraction with graph convolutional layers for global stru...

Full description

Saved in:
Bibliographic Details
Main Authors: Hongliang Zhu, Hongxi Zhao, Chunshan Bao, Yiran Shi, Wenchao He
Format: Article
Language:English
Published: MDPI AG 2025-07-01
Series:Sensors
Subjects:
Online Access:https://www.mdpi.com/1424-8220/25/15/4563
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849770609524867072
author Hongliang Zhu
Hongxi Zhao
Chunshan Bao
Yiran Shi
Wenchao He
author_facet Hongliang Zhu
Hongxi Zhao
Chunshan Bao
Yiran Shi
Wenchao He
author_sort Hongliang Zhu
collection DOAJ
description We propose a hybrid Convolutional Graph Neural Network (C-GNN) for direction-of-arrival (DOA) estimation in sparse sensor arrays under low-snapshot conditions. The C-GNN architecture combines 1D convolutional layers for local spatial feature extraction with graph convolutional layers for global structural learning, effectively capturing both fine-grained and long-range array dependencies. Leveraging the difference coarray technique, the sparse array is transformed into a virtual uniform linear array (VULA) to enrich the spatial sampling; real-valued covariance matrices derived from the array measurements are used as the network’s input features. A final multi-layer perceptron (MLP) regression module then maps the learned representations to continuous DOA angle estimates. This approach capitalizes on the increased degrees of freedom offered by the virtual array while inherently incorporating the array’s geometric relationships via graph-based learning. The proposed C-GNN demonstrates robust performance in noisy, low-data scenarios, reliably estimating source angles even with very limited snapshots. By focusing on methodological innovation rather than bespoke architectural tuning, the framework shows promise for data-efficient DOA estimation in challenging practical conditions.
format Article
id doaj-art-1c501997ca754ac8b34cf18bf1b18f5a
institution DOAJ
issn 1424-8220
language English
publishDate 2025-07-01
publisher MDPI AG
record_format Article
series Sensors
spelling doaj-art-1c501997ca754ac8b34cf18bf1b18f5a2025-08-20T03:02:56ZengMDPI AGSensors1424-82202025-07-012515456310.3390/s25154563Robust Low-Snapshot DOA Estimation for Sparse Arrays via a Hybrid Convolutional Graph Neural NetworkHongliang Zhu0Hongxi Zhao1Chunshan Bao2Yiran Shi3Wenchao He4College of Communications Engineering, Jilin University, Changchun 130015, ChinaCollege of Communications Engineering, Jilin University, Changchun 130015, ChinaCollege of Communications Engineering, Jilin University, Changchun 130015, ChinaCollege of Communications Engineering, Jilin University, Changchun 130015, ChinaCollege of Communications Engineering, Jilin University, Changchun 130015, ChinaWe propose a hybrid Convolutional Graph Neural Network (C-GNN) for direction-of-arrival (DOA) estimation in sparse sensor arrays under low-snapshot conditions. The C-GNN architecture combines 1D convolutional layers for local spatial feature extraction with graph convolutional layers for global structural learning, effectively capturing both fine-grained and long-range array dependencies. Leveraging the difference coarray technique, the sparse array is transformed into a virtual uniform linear array (VULA) to enrich the spatial sampling; real-valued covariance matrices derived from the array measurements are used as the network’s input features. A final multi-layer perceptron (MLP) regression module then maps the learned representations to continuous DOA angle estimates. This approach capitalizes on the increased degrees of freedom offered by the virtual array while inherently incorporating the array’s geometric relationships via graph-based learning. The proposed C-GNN demonstrates robust performance in noisy, low-data scenarios, reliably estimating source angles even with very limited snapshots. By focusing on methodological innovation rather than bespoke architectural tuning, the framework shows promise for data-efficient DOA estimation in challenging practical conditions.https://www.mdpi.com/1424-8220/25/15/4563direction-of-arrival (DOA) estimationgraph neural networklow snapshotsarray signal processing
spellingShingle Hongliang Zhu
Hongxi Zhao
Chunshan Bao
Yiran Shi
Wenchao He
Robust Low-Snapshot DOA Estimation for Sparse Arrays via a Hybrid Convolutional Graph Neural Network
Sensors
direction-of-arrival (DOA) estimation
graph neural network
low snapshots
array signal processing
title Robust Low-Snapshot DOA Estimation for Sparse Arrays via a Hybrid Convolutional Graph Neural Network
title_full Robust Low-Snapshot DOA Estimation for Sparse Arrays via a Hybrid Convolutional Graph Neural Network
title_fullStr Robust Low-Snapshot DOA Estimation for Sparse Arrays via a Hybrid Convolutional Graph Neural Network
title_full_unstemmed Robust Low-Snapshot DOA Estimation for Sparse Arrays via a Hybrid Convolutional Graph Neural Network
title_short Robust Low-Snapshot DOA Estimation for Sparse Arrays via a Hybrid Convolutional Graph Neural Network
title_sort robust low snapshot doa estimation for sparse arrays via a hybrid convolutional graph neural network
topic direction-of-arrival (DOA) estimation
graph neural network
low snapshots
array signal processing
url https://www.mdpi.com/1424-8220/25/15/4563
work_keys_str_mv AT hongliangzhu robustlowsnapshotdoaestimationforsparsearraysviaahybridconvolutionalgraphneuralnetwork
AT hongxizhao robustlowsnapshotdoaestimationforsparsearraysviaahybridconvolutionalgraphneuralnetwork
AT chunshanbao robustlowsnapshotdoaestimationforsparsearraysviaahybridconvolutionalgraphneuralnetwork
AT yiranshi robustlowsnapshotdoaestimationforsparsearraysviaahybridconvolutionalgraphneuralnetwork
AT wenchaohe robustlowsnapshotdoaestimationforsparsearraysviaahybridconvolutionalgraphneuralnetwork