基于Res2Net-ECA的雷达辐射源个体识别  

Radar Specific Emitter Identification Based on Res 2 Net-ECA

作　　者：安林 张文旭[1] 孙富礼 AN Lin;ZHANG Wenxu;SUN Fuli(School of Information and Communication Engineering,Harbin Engineering University,Harbin 150001,Heilongjiang,China;Shanghai Radio Equipment Research Institute,Shanghai 201109,China)

机构地区：[1]哈尔滨工程大学信息与通信工程学院,黑龙江哈尔滨150001 [2]上海无线电设备研究所,上海201109

出　　处：《制导与引信》2024年第4期14-19,38,共7页Guidance & Fuze

基　　金：航空科学基金(202200200P6002);黑龙江省“优秀青年教师基础研究支持计划”重点项目(YQJH2023279)。

摘　　要：针对雷达辐射源个体识别过程中存在的模型泛化、特征表征不足和低信噪比信号识别难度大等问题,提出了一种基于高效通道注意力多尺度残差神经网络(Res2Net-ECA)的雷达辐射源个体识别方法。该方法将高效通道注意力(effective channel attention,ECA)和多尺度残差神经网络(multi-scale residual neural network,Res2Net)相结合,首先对信号进行双谱变换,并利用围线积分对数据进行降维,以降低后续计算的复杂度。然后,利用Res2Net提取数据的多尺度特征,确保数据特征被有效挖掘。最后,引入注意力机制,调整特征通道的权重以突出重要特征,从而进一步提高识别准确率。实验结果表明,该方法能够在不同的信噪比条件下保持较高的识别准确率,具有广阔的应用前景和很大的开发潜力。A radar specific emitter identification method based on Res2Net-ECA was proposed to address the problems of model generalization,insufficient feature representation,and difficulty in identifying low signal-to-noise ratio signal in the process of radar specific emitter identification.This method combined effective channel attention(ECA)with multi-scale residual neural network(Res2Net).Firstly,bispectral transformation was performed on the signal and surround integral was used for reducing the dimensionality of the data,in order to reduce the complexity of subsequent calculations.Then,Res2Net was used for extracting multi-scale features of the data,ensuring that the data features were effectively mined.Finally,an attention mechanism was introduced to adjust the weights of feature channels and highlight important features,thereby further improving the identification accuracy.The experimental results show that the method can maintain high identification accuracy under different signal-to-noise ratio conditions,and has broad application prospect and great development potential.

关 键 词：深度学习 雷达辐射源个体识别 注意力机制 双谱 多尺度特征 

分 类 号：TN957.52[电子电信—信号与信息处理]