基于深度学习的OFDM信号调制方式识别  

Deep Learning Based Signal Modulation Identification in OFDM Systems

作　　者：任小平 眭超亚 蔡小莉 刘小勇 REN Xiaoping;SUI Chaoya;CAI Xiaoli;LIU Xiaoyong(Chongqing Chemical Industry Vocational College,Chongqing 401228,China)

机构地区：[1]重庆化工职业学院,重庆401228

出　　处：《信息与电脑》2023年第18期137-140,共4页Information & Computer

基　　金：重庆市教委科学技术研究项目(项目编号:KJQN202104503);重庆化工职业学院校级科研课题(项目编号:HZY2021-KJ04)。

摘　　要：正交频分复用(Orthogonal Frequency Division Multiplexing,OFDM)技术广泛应用在各类通信系统中。信号调制方式识别对于OFDM技术十分重要,特别是在非合作通信系统中。传统的信号调制方式识别精度不高,且没有研究信噪比为0 dB以下的情况。对此,将OFDM信号看成2×N的图像,保存I、Q两路信号在空间上的特性,设计包含3个卷积层、3个全连接层的深度卷积神经网络(Convolutional Neural Networks,CNN)。将收敛后的模型用于调制方式分类,实验结果表明,所设计的深度神经网络对6种OFDM信号具有很好的识别效果。Orthogonal Frequency Division Multiplexing(OFDM)technology is widely used in various communication systems.Signal modulation mode recognition is very important for OFDM technology,especially in non-cooperative communication systems.The recognition accuracy of the traditional signal modulation method is not high,and the case of signal-to-noise ratio below 0 dB is not studied.In this regard,this paper regards OFDM signals as 2×N images,preserves the spatial characteristics of the I and Q signals,and designs a deep Convolutional Neural Network(CNN)that includes 3 convolutional layers and 3 fully connected layers.The convergent model is applied to modulation classification.Experimental results show that the designed deep neural network has a good recognition effect on 6 kinds of OFDM signals.

关 键 词：正交频分复用(OFDM) 调制方式识别 卷积神经网络(CNN) 深度学习 

分 类 号：TN929.5[电子电信—通信与信息系统]