面向混合非高斯噪声环境的相干信源鲁棒DOA估计方法

王艺泽; 刘磊

doi:10.11999/JEIT251383

面向混合非高斯噪声环境的相干信源鲁棒DOA估计方法

doi: 10.11999/JEIT251383 cstr: 32379.14.JEIT251383

王艺泽^{1, 2, 3},
刘磊^{1, 2, 3, ,}

1.
新疆大学计算机科学与技术学院，新疆大学乌鲁木齐 830046
2.
新疆信号检测与处理重点实验室，乌鲁木齐新疆 830046
3.
丝绸之路多语种认知计算联合国际研究实验室，乌鲁木齐新疆 830046

基金项目: 新疆维吾尔自治区自然科学基金面上项目(项目编号：2023D01C18)，天池人才计划(领军人才)第二批项目

详细信息

作者简介:
王艺泽：男，研究生，研究方向为阵列信号处理的波达方向估计

刘磊：男，博士，副教授，硕导，研究方向为阵列信号处理的波达方向估计，智能算法等

通讯作者:
刘磊　xjuliu@xju.edu.cn

计量
- 文章访问数: 24
- HTML全文浏览量: 14
- PDF下载量: 5
- 被引次数: 0
出版历程
- 修回日期: 2026-03-16
- 录用日期: 2026-03-16
- 网络出版日期: 2026-06-25

Robust DOA Estimation of Coherent Sources in Mixed Non-Gaussian Noise Environments

WANG Yize^{1, 2, 3},
LIU Lei^{1, 2, 3
, ,}

1.
School of Computer Science and Technology, Xinjiang University, Urumgi, Xinjiang 830046, China
2.
Xinjiang Key Laboratory of Signal Detection and Processing, Urumgi, Xinjiang 830046, China
3.
Joint International Research Laboratory of Silk Road Multilingual Cognitive Computing, Urumgi, Xinjiang 830046, China

Funds: Xinjiang Uygur Autonomous Region Natural Science Foundation General Program (Project Number: 2023D01C18), the second batch of Tianchi Talents (Leading Talents) projectin Xinjiang Uygur Autonomous Region

摘要

摘要: 波达方向(DOA)估计是阵列信号处理的基础问题。针对混合非高斯噪声导致的传统子空间算法失效及深度网络在统计特性失配下的抗脉冲干扰能力不足难题，相干信源导致的协方差矩阵秩亏问题，引入前后向空间平滑(FBSS)算法构建秩恢复的预处理机制。提出基于最大相关熵准则(MCC)的多通道特征融合网络(MCFCF)。框架设计了多尺度特征提取(MSFE)与空频注意力融合(SFAF)模块，通过空频域联合建模实现特征的物理校准与增强；采用MCC替代均方误差损失，构建抗脉冲离群值的鲁棒优化目标。实验表明，MCFCF在低信噪比与相干信源环境下的估计精度显著优于MUSIC、ESPRIT及TSPCNN、IQResNet等模型。这一方法为噪声环境中的目标定位与智能信号处理提供了新思路。
- DOA估计 /
- 深度学习 /
- 多尺度特征 /
- 混合注意力机制
Abstract: Objective To resolve DOA estimation failures under mixed non-Gaussian noise, a Multi-Channel Feature Fusion Network (MCFCF) based on the Maximum Correntropy Criterion (MCC) is proposed. By integrating Multi-Scale Feature Extraction (MSFE) and Spatial-Frequency Attention Fusion (SFAF), the framework performs joint spatial-frequency calibration. MCC replaces the Mean Squared Error (MSE) to enhance resilience against impulsive outliers. Experiments show that MCFCF significantly outperforms MUSIC, ESPRIT, and SOTA deep networks in accuracy and super-resolution, particularly under low SNR and coherent source conditions. Methods To resolve DOA estimation failures in mixed non-Gaussian noise and coherent source environments, an MCFCF framework based on the Maximum Correntropy Criterion (MCC) is proposed. The method utilizes FBSS for rank recovery, integrates multi-scale feature extraction (MSFE) with spatial-frequency attention (SFAF) for physical manifold calibration, and adopts the MCC loss function to establish a statistically robust optimization objective against impulsive outliers. Results and Discussions Experimental analysis verifies the superior performance of MCFCF in complex environments (Figs. 5-8). Under mixed noise with strong impulsive and heavy-tailed components, traditional subspace methods fail (RMSE $ \gt {10}^{\circ } $) due to “subspace swapping,” whereas MCFCF maintains a robust $ {1.5}^{\circ } $ RMSE at –15 dB, significantly outperforming SOTA models (Fig. 5). Adaptability tests across heterogeneous distributions (Fig. 6) demonstrate exceptional distribution invariance; in white noise benchmarks, MCFCF compensates for FBSS-induced aperture loss and finite-sample variance, surpassing MUSIC in the low-SNR regime. Mechanism analysis (Figs. 7-8) reveals that MSFE’s multi-scale receptive fields and SFAF’s amplitude-phase decoupling provide the foundation for stability and precision, respectively. Furthermore, the MCC-based "soft-truncation" mechanism fundamentally blocks GCLT-governed impulsive interference at the optimization level, ensuring global robustness. Conclusions An MCC-based MCFCF framework is proposed to resolve DOA estimation failures in mixed non-Gaussian and coherent source environments. By employing FBSS for rank recovery and integrating MSFE with SFAF for manifold calibration, the model adopts an MCC loss to suppress impulsive outliers. Results confirm superior accuracy and distribution invariance at -15 dB SNR, significantly outperforming MUSIC, ESPRIT, and state-of-the-art deep learning baselines while breaching performance lower bounds in hostile environments.
- DOA estimation /
- Deep learning /
- Multi-scale features /
- Hybrid attention mechanism

HTML全文

图 1 MCFCF网络架构

下载: 全尺寸图片幻灯片

图 2 Multi-scale feature extraction(MSFE)

下载: 全尺寸图片幻灯片

图 3 Spatial frequency attention fusion(SFAF)

下载: 全尺寸图片幻灯片

图 4 多源特征有机动态融合与骨干网络

下载: 全尺寸图片幻灯片

图 5 各算法在混合噪声下的性能对比

下载: 全尺寸图片幻灯片

图 6 各算法在单一主导噪声下的表现

下载: 全尺寸图片幻灯片

图 7 消融研究结果

下载: 全尺寸图片幻灯片

图 8 MCC vs MSE

下载: 全尺寸图片幻灯片

参考文献(22)

[1]	康伟泉, 鲁祖坤, 李柏渝, 等. 面向欺骗抑制的GNSS信号单通道高精度稀疏DOA估计[J]. 电子与信息学报, 2025, 47(12): 4917–4925. doi: 10.11999/JEIT250725. KANG Weiquan, LU Zukun, LI Baiyu, et al. Single-channel high-precision sparse DOA estimation of GNSS signals for deception suppression[J]. Journal of Electronics & Information Technology, 2025, 47(12): 4917–4925. doi: 10.11999/JEIT250725.
[2]	WANG Zhuying, YAN Yongsheng, HE Ke, et al. Augmented matrix construction based DOA estimation of coherent underwater acoustic signals[C]. 2023 IEEE International Conference on Signal Processing, Communications and Computing, Zhengzhou, China, 2023: 1–5. doi: 10.1109/ICSPCC59353.2023.10400250.
[3]	陈涛, 席豪林, 詹磊, 等. 基于FRIDA的多通道切换阵列DOA估计算法[J]. 电子与信息学报, 2025, 47(12): 5106–5115. doi: 10.11999/JEIT250350. CHEN Tao, XI Haolin, ZHAN Lei, et al. Multi-channel switching array DOA estimation algorithm based on FRIDA[J]. Journal of Electronics & Information Technology, 2025, 47(12): 5106–5115. doi: 10.11999/JEIT250350.
[4]	邢传玺, 黄廷龙, 谈光枝, 等. 融合空间域维纳滤波与卷积神经网络的水声DOA估计[J]. 电子与信息学报, 2025, 47(10): 3735–3744. doi: 10.11999/JEIT250141. XING Chuanxi, HUANG Tinglong, TAN Guangzhi, et al. Acoustic DOA estimation in underwater environments by integrating spatial domain wiener filtering and convolutional neural networks[J]. Journal of Electronics & Information Technology, 2025, 47(10): 3735–3744. doi: 10.11999/JEIT250141.
[5]	BARTLETT M S. Properties of sufficiency and statistical tests[J]. Proceedings of the Royal Society of London. A. Mathematical and Physical Sciences, 1937, 160(901): 268–282. doi: 10.1098/rspa.1937.0109.
[6]	CAPON J. High-resolution frequency-wavenumber spectrum analysis[J]. Proceedings of the IEEE, 1969, 57(8): 1408–1418. doi: 10.1109/PROC.1969.7278.
[7]	SCHMIDT R. Multiple emitter location and signal parameter estimation[J]. IEEE Transactions on Antennas and Propagation, 1986, 34(3): 276–280. doi: 10.1109/TAP.1986.1143830.
[8]	ROY R, PAULRAJ A, and KAILATH T. Estimation of signal parameters via rotational invariance techniques – ESPRIT[C]. MILCOM 1986 - IEEE Military Communications Conference: Communications-Computers: Teamed for the 90’s, Monterey, USA, 1986: 41.6. 1–41.6. 5. doi: 10.1109/MILCOM.1986.4805850.
[9]	LIU Fulai, LI Xubin, SU Zhibo, et al. Mc-music algorithm for fast DOA estimation[J]. Wireless Networks, 2025, 31(1): 659–667. doi: 10.1007/s11276-024-03794-2.
[10]	WANG Rui, WANG Yue, LI Yanping, et al. Geometric algebra-based ESPRIT algorithm for DOA estimation[J]. Sensors, 2021, 21(17): 5933. doi: 10.3390/s21175933.
[11]	ZHOU Haocheng, LIU Zhenghong, LUO Liyan, et al. An improved two-stage spherical harmonic ESPRIT-type algorithm[J]. Symmetry, 2023, 15(8): 1607. doi: 10.3390/sym15081607.
[12]	LV Weihua, SUN Huapu, ZHANG Xiaofei, et al. Reduced-dimension noncircular-capon algorithm for DOA estimation of noncircular signals[J]. International Journal of Antennas and Propagation, 2015, 2015: 348636. doi: 10.1155/2015/348636.
[13]	LIU Zhangmeng, ZHANG Chenwei, and YU P S. Direction-of-arrival estimation based on deep neural networks with robustness to array imperfections[J]. IEEE Transactions on Antennas and Propagation, 2018, 66(12): 7315–7327. doi: 10.1109/TAP.2018.2874430.
[14]	PAPAGEORGIOU G K, SELLATHURAI M, and ELDAR Y C. Deep networks for direction-of-arrival estimation in low SNR[J]. IEEE Transactions on Signal Processing, 2021, 69: 3714–3729. doi: 10.1109/TSP.2021.3089927.
[15]	ZHENG Shilian, YANG Zhuang, SHEN Weiguo, et al. Deep learning-based DOA estimation[J]. IEEE Transactions on Cognitive Communications and Networking, 2024, 10(3): 819–835. doi: 10.1109/TCCN.2024.3360527.
[16]	LI Xiaoqiang, CHEN Jianfeng, BAI Jisheng, et al. Deep learning-based DOA estimation using CRNN for underwater acoustic arrays[J]. Frontiers in Marine Science, 2022, 9: 1027830. doi: 10.3389/fmars.2022.1027830.
[17]	YU Hao, WU Sheng, LV Liujie, et al. A robust super-resolution DoA estimation algorithm using vision transformer[C]. 2024 13th International Conference on Communications, Circuits and Systems, Xiamen, China, 2024: 413–417. doi: 10.1109/ICCCAS62034.2024.10652636.
[18]	SHMUEL D H, MERKOFER J P, REVACH G, et al. SubspaceNet: Deep learning-aided subspace methods for DoA estimation[J]. IEEE Transactions on Vehicular Technology, 2025, 74(3): 4962–4976. doi: 10.1109/TVT.2024.3496119.
[19]	LI Ninghui, ZHANG Xiaokuan, LV Fan, et al. A Bayesian deep unfolded network for the off-grid direction-of-arrival estimation via a minimum hole array[J]. Electronics, 2024, 13(11): 2139. doi: 10.3390/electronics13112139.
[20]	WU Liuli, LIU Zhangmeng, and HUANG Zhitao. Deep convolution network for direction of arrival estimation with sparse prior[J]. IEEE Signal Processing Letters, 2019, 26(11): 1688–1692. doi: 10.1109/LSP.2019.2945115.
[21]	KONG Lingshun, DONG Jiangxin, GE Jianjun, et al. Efficient frequency domain-based transformers for high-quality image deblurring[C]. 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Vancouver, Canada, 2023: 5886–5895. doi: 10.1109/CVPR52729.2023.00570.
[22]	TOFIGHI M, GUO Tiantong, VANAMALA J K P, et al. Prior information guided regularized deep learning for cell nucleus detection[J]. IEEE Transactions on Medical Imaging, 2019, 38(9): 2047–2058. doi: 10.1109/TMI.2019.2895318.