Research On Signal Modulation Identification Technology Based On Deep Learning In OFDM Systems

Modulation recognition is widely used in the civil and military fields.Especially in military scenarios,electronic warfare,surveillance,and threat analysis all need to first identify the modulation method of the signal,then recognize the enemy's sending unit,and finally interfere with the signal and restore the intercepted signal.In Orthogonal Frequency Division Multiplexing(OFDM)systems,signal modulation recognition is an indispensable and challenging problem.Modulation recognition can be divided into recognition based on likelihood ratio and recognition based on signal features.The former is computationally complicated,and the latter is implemented based on feature extraction and classifiers.However,the traditional methods based on high-order cumulants and signal instantaneous characteristics have limited recognition accuracy under OFDM systems.The rapid development of deep learning has promoted continuous progress in computer vision and natural language processing.A modulation recognition method based on deep learning is proposed,using neural networks to replace feature extraction and classifiers in traditional methods to further improve recognition performance.This paper proposes a modulation recognition technology based on convolutional neural network.Experimental analysis shows that the functions of feature extraction and classification can be achieved simultaneously through convolutional neural networks.The simulation results show that the modulation recognition technology based on convolutional neural network has higher recognition accuracy and classification consistency,and its performance in feature extraction is better than traditional methods.In addition,this paper studies the performance of the model on two different data sets by increasing signal diversity.The simulation results show that the modulation recognition model based on the convolutional neural network has high robustness.Considering that the modulation recognition technology based on convolutional neural network has high computational complexity and model size,it is difficult to be widely used in edge devices.Inspired by Shuffle Net,this paper proposes a modulation recognition technology based on a lightweight neural network.The network model is constructed by combining three methods of depthwise convolution,channel shuffle and global pooling.Theoretical analysis and simulation results show that the Shuffle Net based modulation recognition method has lower computational complexity,fewer model parameters,and smaller model size while ensuring recognition performance.At the same time,the7)₂ regularization and Dropout layer are introduced to prevent overfitting and speed up the training of the model.