| Signal modulation recognition is widely used in electronic reconnaissance equipment,which is of great significance and practical value to improve the technical ability and level of electronic warfare.In the current information battlefield,the number of various radar,communication and other radiation sources is increasing.Modulation forms are becoming more and more diverse.The increasing signal density makes the electromagnetic environment more and more complex.How to identify multi-component signals accurately and quickly from such a complex,dense and changeable electromagnetic environment is an urgent problem to be solved.Aiming at the problems of low accuracy,poor robustness and poor real-time performance of existing modulation recognition methods for multi-component signal recognition in complex electromagnetic environment,this paper puts forward corresponding solutions.The main research contents of this paper are as follows:To address the low recognition rate of traditional feature extraction classification methods for modulated signals in practical communication environments,a feature-based graph generation signal recognition method is proposed in this study.By extracting the instantaneous features and higher-order spectral features of the modulated signals as joint features to construct the node features of the signal,and using the Pearson correlation coefficient to establish edge relationships based on a threshold to complete the graph domain transformation of the modulated signals.Finally,the graph neural network model is constructed to aggregate strongly correlated node features in the modulated signal graph based on the topological relationship.Experimental results on the RML2016.10 a dataset show that the proposed feature graph generation method combined with the constructed graph neural network has higher accuracy than the existing joint feature input support vector machine or K-nearest neighbor classifier recognition methods.The recognition accuracy of the Diff Pool graph neural network is 80.59%at SNR 4d B and threshold 0.8,which proves that the proposed method can further aggregate related features on the basis of feature extraction to improve recognition accuracy.Although the feature-based graph generation signal recognition method surpasses the original feature extraction recognition method,it is still affected by the poor robustness of traditional feature extraction methods for practical communication signals,making it difficult to achieve optimal results.In this study,modulated signals are mapped to the graph domain using the visual graph and horizontal visual graph.To address the problems of fixed mapping rules and low recognition efficiency of the visual graph method,adaptive visible graph generation algorithms based on dual-channel convolution and complex convolution are proposed respectively on the basis of the adaptive graph generation algorithm.Together with the improved Diff Pool graph neural network,they constitute the end-to-end recognition model for modulated signals.The proposed recognition model is validated on the RML2016.10 a dataset and compared with multiple existing recognition models.The experimental results show that the graph generation recognition method based on complex convolution has a recognition accuracy of 92.05% at SNR 10 d B,which is superior to the feature-based graph generation signal recognition method and the models used for comparison. |
