基于Bagging-CNN雷达信号分类方法  

Radar Signal Classification Method Based on Bagging-CNN

作　　者：孙艺聪 田润澜[1] 刘冲 郭扬[3] SUN Yicong;TIAN Runlan;LIU Chong;GUO Yang(School of Aviation Operations and Services, Aviation University of Air Force, Changchun 130022, China;Air Force Staff Electronic Warfare Radar Bureau, Beijing 100000, China;Air Force Aviation University Education support Center, Changchun 130022, China)

机构地区：[1]空军航空大学航空作战勤务学院,长春130022 [2]空军参谋部电子对抗雷达局,北京100000 [3]空军航空大学教研保障中心,长春130022

出　　处：《兵器装备工程学报》2021年第5期191-195,226,共6页Journal of Ordnance Equipment Engineering

基　　金：国家自然科学基金项目(61571462)。

摘　　要：针对雷达辐射源信号在低信噪比情况下分类精度不高、分类结果具有偏向性,设计了一种基于集成学习结合深度学习的雷达信号分类方法。首先利用模糊函数特征对雷达信号进行特征提取,对提取的图像特征进行预处理得到网络训练的数据集。然后构造一个集成多个卷积神经网络分类模型对雷达信号进行识别。采用5种不同的雷达信号进行验证,实验结果表明,相比于单个卷积神经网络,采用集成学习和卷积神经网络结合的分类方法有助于提高低信噪比信号的分类效果。In order to solve the problem of low classification accuracy and biased classification results of radar radiation signals at low SNR,a radar signal classification method based on integrated learning and deep learning was designed.The fuzzy function feature was used to extract the radar signal,and the extracted image feature was processed to obtain the network training data set.Then an integrated multiple conventional neural network classification model is constructed to identify radar signals.Five different radar signals are used for verification.The experimental results show that compared with a single conventional neural network,the classification method combining integrated learning and conventional neural network is helpful to improve the classification effect of low SNR signals.

关 键 词：集成学习 卷积神经网络 模糊函数 图像预处理 信号分类 

分 类 号：TN971.+1[电子电信—信号与信息处理]