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

Area of Science:

• Signal Processing
• Machine Learning
• Array Signal Processing

Background:

• Direction-of-arrival (DOA) estimation is crucial for applications like radar and wireless communications.
• Sparse sensor arrays and low-snapshot conditions present significant challenges for traditional DOA estimation methods.
• Existing techniques often struggle with reduced spatial sampling and limited data availability.

Purpose of the Study:

• To develop a robust and data-efficient DOA estimation method for sparse sensor arrays under low-snapshot conditions.
• To leverage hybrid deep learning architectures for enhanced feature extraction and structural learning in DOA estimation.
• To improve the accuracy and reliability of DOA estimation in challenging, real-world scenarios.

Main Methods:

• A hybrid Convolutional Graph Neural Network (C-GNN) architecture is proposed, integrating 1D convolutional layers and graph convolutional layers.
• The difference coarray technique is employed to transform the sparse array into a virtual uniform linear array (VULA), increasing degrees of freedom.
• Real-valued covariance matrices from array measurements serve as input features, processed by an MLP regression module for continuous DOA estimation.

Main Results:

• The C-GNN effectively extracts local spatial features and learns global structural dependencies from array data.
• The virtual array technique enriches spatial sampling, enhancing the network's ability to capture signal information.
• The proposed method demonstrates robust performance in noisy, low-snapshot environments, achieving reliable DOA estimation.

Conclusions:

• The hybrid C-GNN offers a promising, data-efficient approach for DOA estimation in sparse arrays with limited snapshots.
• The integration of convolutional and graph neural networks provides a powerful framework for exploiting array geometry and signal characteristics.
• This methodology shows potential for practical applications facing challenging signal acquisition constraints.