Research On Modulation Recognition Method Of Electromagnetic Signal Based On Deep Learning

In recent years,with the rapid development of wireless communication,the internet of things,artificial intelligence and other emerging technologies,the number of electromagnetic equipment has grown exponentially,resulting in an increasingly complex electromagnetic environment.Electromagnetic signal detection and recognition technology is faced with many challenges such as dynamic electromagnetic environment,increasingly diverse data sources and rapidly increasing processing requirements.As the key link of electromagnetic signal detection and recognition,electromagnetic signal modulation recognition technology is bearing the brunt of the new impact.Traditional recognition methods of electromagnetic signal modulation mainly rely on manual extraction of features,which have poor feature quality and low recognition accuracy.With the deep integration of artificial intelligence and wireless communication technology,signal modulation recognition based on deep learning has attracted much attention.However,the intelligent identification method of electromagnetic signal modulation mode often relies on single domain signal data,and the cost of electromagnetic data acquisition and annotation is high,which leads to the limitation of the identification accuracy of the modulation mode.For the electromagnetic signal modulation mode intelligent recognition technology bottleneck,this topic focus on electromagnetic signal modulation mode recognition method based on the deep learning,based on multiple domain feature fusion is proposed and based on multisource feature migration way of signal modulation recognition methods,design the electromagnetic signal modulation mode recognition system,realized the electromagnetic signal modulation mode of rapid and accurate identification.The main work and innovations of this project are summarized as follows:(1)In the aspect of cross-domain fusion of electromagnetic signal features,a signal modulation recognition method based on multi-domain feature fusion is proposed to solve the problem that most of the current modulation recognition methods lack the complementarity and fusion of electromagnetic signal features in different domains.Firstly,features are extracted from time domain signals by sequence learning model,and features are extracted from time frequency diagram of electromagnetic signals by residual network model.Then the feature information of these different dimensions is fused by the gate mechanism network.Finally,the fusion eigenvectors are used to classify signal modulation modes.Experimental results show that the proposed method is superior to the modulation recognition method based on single domain characteristic information.(2)In the aspect of cross-domain transfer of electromagnetic signal features,a signal modulation mode recognition method based on multisource feature transfer is proposed to solve the problem that the current modulation mode recognition methods based on transfer learning mainly focus on the single source domain and the single target domain.Firstly,a shared feature extractor is used to map data from different source and target domains to the same feature space.Then modulation mode classifier and domain classifier are used to identify the modulation mode and the domain of the electromagnetic signal.Finally,the trained shared feature and modulation classifier are used for modulation classification.Experimental results show that the proposed method is superior to the modulation recognition method based on single source domain feature transfer.(3)Using PyQt5 design and implementation of electromagnetic signal modulation recognition system.Firstly,the system receives the electromagnetic signal collected by Universal Software Radio Peripheral(USRP)in real time through socket transmission protocol,and stores the electromagnetic signal data into MongoDB database.Then the collected electromagnetic signal is preprocessed and the processed data is input into the classification model based on multi-domain feature fusion for real-time online inference recognition.Finally,the results of recognition and the information of electromagnetic signal in different dimensions are displayed in real time.