(SP)$^2$-Net: A Neural Spatial Spectrum Method for DOA Estimation

2025-09-18

AI Summary

Direction-of-arrival (DOA) estimation under single-snapshot, unknown source number, and closely spaced source conditions remains challenging; conventional Bartlett beamformers suffer from limited physical array aperture, resulting in insufficient resolution and accuracy. Method: This paper proposes a deep learning–based high-resolution spatial spectrum estimation framework. A dedicated neural network is designed to take a single array snapshot and candidate angle hypotheses as input, enabling end-to-end learning of a super-resolved spatial spectrum. Through an innovative architecture and training strategy, the model implicitly learns an equivalent extended array, thereby surpassing the Rayleigh limit. Inference requires only angular scanning to generate a full-field spatial heatmap. Contribution/Results: Experiments demonstrate that the proposed method significantly outperforms Bartlett beamforming and state-of-the-art sparse recovery algorithms under single-snapshot conditions, achieving superior estimation accuracy, enhanced robustness to noise and model mismatch, and substantially improved angular resolution.

Technology Category

• Intelligent Robots: State Estimation
• Machine Learning: Deep Learning Algorithms
• Search and Optimization: Learning to Search

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

We consider the problem of estimating the directions of arrival (DOAs) of multiple sources from a single snapshot of an antenna array, a task with many practical applications. In such settings, the classical Bartlett beamformer is commonly used, as maximum likelihood estimation becomes impractical when the number of sources is unknown or large, and spectral methods based on the sample covariance are not applicable due to the lack of multiple snapshots. However, the accuracy and resolution of the Bartlett beamformer are fundamentally limited by the array aperture. In this paper, we propose a deep learning technique, comprising a novel architecture and training strategy, for generating a high-resolution spatial spectrum from a single snapshot. Specifically, we train a deep neural network that takes the measurements and a hypothesis angle as input and learns to output a score consistent with the capabilities of a much wider array. At inference time, a heatmap can be produced by scanning an arbitrary set of angles. We demonstrate the advantages of our trained model, named (SP)$^2$-Net, over the Bartlett beamformer and sparsity-based DOA estimation methods.

Problem

• Estimating directions of arrival from single snapshot measurements
• Overcoming resolution limitations of classical Bartlett beamformer
• Generating high-resolution spatial spectrum using neural network

Innovation

• Deep neural network for single snapshot DOA
• Inputs measurements and hypothesis angle
• Outputs high-resolution spatial spectrum