基于阈值去噪神经网络的干扰用频模式识别  

Jamming Frequency Pattern Recognition Based on Threshold Denoising Neural Network

作　　者：姚昌华 李洋 YAO Changhua;LI Yang(School of Electronics and Information Engineering,Nanjing University of Information Science and Technology,Nanjing 210044,China)

机构地区：[1]南京信息工程大学电子与信息工程学院,南京210044

出　　处：《电子信息对抗技术》2025年第2期1-7,共7页Electronic Information Warfare Technology

基　　金：通信抗干扰全国重点实验室基础科研创新基金(稳定支持)项目资助(IFN20230207);国家自然科学基金资助项目(U22B2002,61971439)。

摘　　要：对电磁干扰模式的准确识别,能为无线通信抗干扰提供重要的决策依据。目前对该问题仅有少量研究,且未考虑环境噪声对其识别性能和方法设计的重要影响。针对噪声条件下干扰模式识别提出了一种基于阈值去噪的深度学习模型,以在特征空间内减弱噪声特征对识别的影响。并分析了去噪功能在不同深度网络层中的效果,提出在卷积神经网络的浅层加入阈值去噪块,以克服噪声对浅层特征的影响,以及非线性传播导致的对深层特征的进一步影响,提高噪声条件下干扰模式识别的准确率。通过对12种不同干扰模式进行识别并对所提方法的性能进行评估,实验结果表明,所提阈值去噪模型和部署优化方法有效降低了噪声对干扰识别的影响,在处理干扰模式识别任务上取得良好的性能。The accurate identification of electromagnetic jamming patterns provides crucial decision-making support for enhancing the anti-jamming capability of wireless communication.There is currently limited research on this issue,and the significant impact of environmental noise on recognition performance and method design has not been adequately considered.Aiming at jamming pattern recognition under noisy conditions,a deep learning model based on automatic updating threshold denoising is proposed.This model aims to mitigate the influence of noise features on recognition within the feature space.Meanwhile,the effectiveness of denoising functions at different depths within the network is analyzed and incorporating threshold denoising blocks are suggested in the shallow layers of convolutional neural networks,in order to overcome the impact of noise on shallow features and further address the effects propagated to deep features due to nonlinearities,enhancing recognition accuracy of jamming patterns under noisy conditions.Performance evaluation is conducted by recognizing 12 different jamming patterns,demonstrating that the proposed threshold denoising model and deployment optimization methods effectively reduce the impact of noise on jamming recognition,and achieve good performance in handling jamming pattern recognition tasks.

关 键 词：通信抗干扰 干扰用频模式识别 阈值去噪 浅层去噪 

分 类 号：TN973.3[电子电信—信号与信息处理]