| 摘要: |
| 针对基于图像识别的雷达信号调制样式识别方法生成图像耗时长，收敛速度慢，且在低信噪比条件下识别准确率低的问题，提出了一种新的雷达信号调制样式识别模型。此模型将雷达信号时间序列经简单预处理后直接作为网络输入，避免了将信号转换为图像的复杂过程；网络能够自主提取雷达信号空间和时间两个方面特征，完善了特征提取的方式；同时，对网络结构进行了优化，并引入了选择性核网络，以获取特征在不同尺度上的重要信息。实验结果表明，此模型在低信噪比条件下具有较快的训练速度和较高的识别准确率。 |
| 关键词: 雷达信号 调制样式识别 深度学习 选择性核网络（SK-Net） 时间序列 |
| DOI：10.20079/j.issn.1001-893x.220216002 |
|
| 基金项目:国家自然科学基金资助项目（61571462） |
|
| Radar signal modulation type recognition based on BSRes_SK_GRU |
| LIU Yuxin,TIAN Runlan,REN Lin,SUN Liang |
| (School of Aviation Operations and Services,Aviation University of Air Force,Changchun 130022,China;Unit 93110 of PLA,Beijing 100843,China) |
| Abstract: |
| For the problems of long time-consuming image generation,slow convergence speed and low recognition accuracy under the condition of low signal-to-noise ratio(SNR),a new model is proposed for radar signal modulation type recognition.In this model,the time series of radar signals are directly used as network input after simple preprocessing,which avoids the complex process of converting signals into images.The network can independently extract the spatial and temporal features of radar signals,which improves the way of feature extraction.At the same time,the network structure is optimized,and the Selective Kernel Network(SK-Net) is introduced to obtain the important information of features at different scales.The experimental results show the proposed model has faster speed and higher accuracy under the condition of low SNR. |
| Key words: radar signal modulation type recognition deep learning selective kernel network(SK Net) time series |
《电讯技术》编辑部 