基于深度学习的单多载波信号识别方法  

Single Multi-carrier Signal Recognition Method based on Deep Learning

作　　者：杜志毅 李保国[1] 张澄安 徐强[1] DU Zhiyi;LI Baoguo;ZHANG Cheng’an;XU Qiang(National University of Defense Technology,Changsha Hunan 410073,China)

机构地区：[1]国防科技大学,湖南长沙410073

出　　处：《通信技术》2021年第5期1070-1075,共6页Communications Technology

基　　金：湖南省自然科学基金创新群体项目“非合作智能信号处理技术”(No.2019JJ10004)。

摘　　要：信号调制识别技术在正交频分复用(Orthogonal Frequency Division Multiplexing,OFDM)系统中有着重要的应用。然而,在单多载波识别方面,大多数传统方法为模型驱动,需要对信号进行预处理。另外其适应信道环境单一,在复杂信道环境时识别性能较差。针对以上问题,提出了一种基于改进ResNet深度学习网络的调制识别方法(Improved ResNet based Modulation Classification method,IRMC)。IRMC通过对同向正交IQ序列进行训练,自主实现了单多载波信号的有效区分。仿真实验表明,单多载波识别方面,IRMC在复杂莱斯多径信道上相比于传统方法在85%识别率下能够达到10 dB的信噪比增益,证明提出的方法是有效可行的。Signal modulation recognition technology has an important application in OFDM(Orthogonal Frequency Division Multiplexing)systems.However,in terms of single-multi-carrier identification,most of the traditional methods are model-driven and require signal preprocessing.In addition,it adapts to single channel environment and has poor recognition performance in complex channel environment.Aim at the above problems,this paper proposes an improved ResNet based modulation classification method(IRMC).By training the IQ(In-phase Quadrature)samples,IRMC realizes the effective distinction between singlecarrier signals and OFDM signals.Experiments indicate that in terms of single-multi-carrier identification,IRMC can achieve a 10 dB SNR gain at 85%recognition rate compared to traditional methods on complex Rice multipath channels,which proves that the method proposed in this paper is effective.

关 键 词：OFDM信号 调制识别 深度学习 ResNet网络 

分 类 号：TN911[电子电信—通信与信息系统]