基于改进ResNet的射频指纹识别方法  

A Radio Frequency Fingerprinting Identification Method Based on Improved ResNet

作　　者：谢跃雷[1,2] 邓涵方 XIE Yuelei;DENG Hanfang(School of Information and Communication,Guilin University of Electronic Technology,Guilin 541004,China;National and Local Joint Engineering Research Centre for Satellite Navigation,Positioning and Location Services,Guilin 541004,China)

机构地区：[1]桂林电子科技大学信息与通信学院,广西桂林541004 [2]卫星导航定位与位置服务国家地方联合工程研究中心,广西桂林541004

出　　处：《电讯技术》2022年第4期416-423,共8页Telecommunication Engineering

基　　金：国家自然科学基金资助项目(6146105);广西科技重大专项(桂科AA21077008)。

摘　　要：在无线网络安全和可能存在的射频设备管理应用范围内,针对多个发射同种射频信号的高度相似射频设备的分类识别问题,提出了一种信号双谱与改进的残差神经网络(Residual Neural Network,ResNet)的射频指纹识别方法。首先,将采集到的不同设备的信号做双谱变换得到双谱等高图并打上设备标签,再使用搭建好的改进残差神经网络模型训练双谱等高图,通过反向传播(Back Propagation,BP)与梯度下降来更新网络权重得到最优化模型,最后使用另外一组双谱等高图验证识别性能。实验结果表明,基于信号双谱与改进的残差神经网络算法在实际电磁环境下识别率达到95.2%,是一种有效的射频指纹分类识别方法。In the range of wireless network security and possible radio frequency(RF)device management applications,for the problem of classification and identification of multiple highly similar RF devices transmitting the same RF signal,an RF fingerprinting identification method based on signal bispectrum and improved residual neural network(ResNet)is proposed.Firstly,the bispectrum contour map is obtained by bispectrum transformation of the collected signals from different devices and labeled with the device label.Then,the bispectrum contour map is trained by the improved ResNet model.The network weight is updated by back propagation(BP)and gradient descent to obtain the optimal model.Finally,another set of bispectrum contour map is used to verify the recognition performance.The experimental results show that the algorithm based on signal bispectrum and improved ResNet has a recognition rate of 95.2%in the actual electromagnetic environment,so it is an effective RF fingerprinting classification and identification method.

关 键 词：射频指纹识别 双谱等高图 深度学习 反向传播 残差神经网络 

分 类 号：TN911.7[电子电信—通信与信息系统] TP183[电子电信—信息与通信工程]