基于多语义特征融合网络的压制干扰识别算法  

Satellite suppression interference identification based on multi-semantic feature fusion network

作　　者：王李军 葛红妨[3] 韩煜 王勃 沈雷[3] WANG Lijun;GE Hongfang;HAN Yu;WANG Bo;SHEN Lei(Key Laboratory of Communication Information Control and Security Technology,Jiaxing Zhejiang 314000,China;The 36th Research Institute of China Electronics Technology Corporation,Jiaxing Zhejiang 314000,China;School of Communication Engineering,Hangzhou Dianzi University,Hangzhou Zhejiang 310018,China)

机构地区：[1]通信信息控制和安全技术重点实验室,浙江嘉兴314000 [2]中国电子科技集团公司第三十六研究所,浙江嘉兴314000 [3]杭州电子科技大学通信工程学院,浙江杭州310018

出　　处：《杭州电子科技大学学报（自然科学版）》2022年第6期1-9,共9页Journal of Hangzhou Dianzi University：Natural Sciences

摘　　要：采用基于时域构造的残差网络干扰识别算法提取压制干扰特征时,泛化能力较弱。为此,提出一种基于多语义特征融合网络的压制干扰识别算法。针对受7类压制干扰影响的卫星信号的识别问题,首先,通过信号变换和填充完成卫星信号的图像化构造;然后,运用跳跃连接的残差网络提取时频域特征和抽象语义特征;最后,通过特征融合模块完成低层和高层输出特征的融合,获得更为复杂和抽象的多语义特征,实现压制干扰信号的识别。实验结果表明,当信噪比大于0 dB或者干信比大于0 dB时,提出算法的识别率均达到90%。Due to the existing residual network interference identification algorithm based on time domain structure has the problem of weak generalization ability of extracting and suppressing interference features,this paper proposes a multi-semantic feature fusion network-based suppression interference identification algorithm.Firstly,the image structure of satellite signals affected by seven kinds of suppressed interference signals is completed through signal transformation and filling;then,the time-frequency domain features and abstract semantic features are extracted by the residual network of skip connections;finally,the low-level features are completed through the feature fusion module.Fusion with high-level output features can obtain more complex and abstract multi-semantic features,and realize the identification of seven kinds of suppressed interference signals suffered by satellite signals.Experimental results show that when the signal-to-noise ratio is greater than 0 dB or the interference-to-signal ratio is greater than 0 dB,the interference recognition rate of the multi-semantic feature fusion network in this thesis can reach 90%.

关 键 词：压制干扰 图像化构造 残差网络 特征融合 

分 类 号：TN972.1[电子电信—信号与信息处理]