Automatic Modulation Recognition of Unknown Interference Signals Based on Graph Model

Automatic Modulation Recognition (AMR), which involves blind identification of interference modulation classes, is an essential technique for maintaining communication security. In this letter, a framework for end-to-end interference sequence recognition is designed by introducing graph signal and graph neural network into the field of interference signals modulation recognition. The framework consists of three modules: channel interaction module, graph mapping and graph feature extraction, and feature fusion classification. The framework focuses, on the one hand, on enhancing the channel information interaction through the proposed channel fusion method and, on the other hand, modelling the sequence as a graph and extracting local features and global features using the graph neural network. Simulation results verify that the method has high performance in the AMR task.