一种改进残差神经网络的调制信号识别方法  

Modulated signal recognition method based on improved residual neural network

作　　者：陈颖 李迟生[1] 张睿琦 CHEN Ying;LI Chisheng;ZHANG Ruiqi(School of Information Engineering,Nanchang University,Nanchang 330031,China)

机构地区：[1]南昌大学信息工程学院,江西南昌330031

出　　处：《现代电子技术》2021年第15期54-58,共5页Modern Electronics Technique

基　　金：国家自然科学基金项目(61661030)。

摘　　要：在无线通信中,调制信号识别分类是非合作通信的重要组成部分,基于特征提取和模式识别的通信信号自动调制识别技术是无线电领域中的重点研究对象,传统的方法无法对各种调制信号进行分类,并且存在计算复杂度高且准确率低的缺点,利用神经网络可以实现多种调制信号的自动识别并取得很好的效果。文中提出一种改进残差神经网络的调制信号识别方法,改进的神经网络通过增大宽度,形成多流结构的残差神经网络,利用多流网络中不同的残差块提取出调制信号的多种特征,进而丰富了提取的信号特征类型,并且避免由于深度加深造成的过拟合现象。仿真实验结果表明,该神经网络框架优于现有的基于深度学习的方法,在信噪比为10 dB时能达到95%以上的准确率,有效地提高了数字调制信号自动识别的准确率。The modulated signal recognition and classification are two important parts of non⁃cooperative communication in wireless communication.The communication signal automatic modulation recognition technology based on feature extraction and pattern recognition is a key research object in the field of radio.The traditional methods are unable to classify various modulated signals and are of high computational complexity and low accuracy.However,the neural network can be used to automatically recognize the multiple modulated signals and has achieved good results.In view of this,a modulated signal recognition method based on improved residual neural network is proposed.The improved neural network can form the residual neural network with multi⁃flow structure by increasing its width,and utilizes the different residual blocks in the multi⁃flow network to extract a variety of features of the modulated signals,so as to enrich the feature types of the extracted signals and avoid over fitting caused by depth deepening.The simulation results show that the neural network framework of the proposed method is better than that of the existing methods based on deep learning,and its accuracy can reach more than 95%when the signal⁃to⁃noise ratio(SNR)is 10 dB.Therefore,the proposed method can effectively improve the accuracy of automatic recognition of digital modulated signals.

关 键 词：调制信号识别 残差神经网络 无线通信 特征提取 模式识别 过拟合 仿真分析 

分 类 号：TN911.3-34[电子电信—通信与信息系统]