Modulation Recognition With Graph Convolutional Network

Yabo Liu, Yi Liu, Cheng Yang

Abstract: In most non-cooperative communication systems, modulation recognition is a fundamental and critical technique. Traditional methods of modulation recognition can be categorized as maximum likelihood hypothesis algorithms and pattern recognition algorithms. However, these methods have high complexities or need additional data preprocessing. Recently, neural network algorithms have shown great potential in modulation recognition. In this letter, we propose a method of modulation recognition by exploiting the graph convolutional network (GCN). However, GCNs cannot be directly used to perform modulation recognition since modulated signals are not graphs. To convert signals to graphs, the modulation dataset is divided into multiple subsets. We design a feature extraction convolutional neural network (CNN) and a graph mapping CNN to extract signal features and map subsets into graphs, respectively. Then we input the graphs into the GCN to predict modulation modes of unlabeled signals. The experimental results show that the proposed GCN algorithm achieves higher recognition accuracy than CNN algorithm and K-nearest neighbor (KNN) algorithm, especially when SNR is low.