When Diffusion Meets AMR: A Multi-Channel Fused Graph Transformer for Automatic Modulation Recognition under Large CFOs and SROs

abstract-As a crucial component of non-cooperative communication, automatic modulation recognition (AMR) has received widespread attention. However, due to the limitation of hardware devices or interference from non-ideal environments, the scale of carrier frequency offset (CFO) and sampling rate offset (SRO) can be significantly large, which seriously damages the recognition accuracy. To address this issue, we propose a novel graph transformer-based method with multi-channel inputs, where each segment of the signal is treated as a node. Inspired by the diffusion process with energy constraint, the network tries to achieve a balance between the representation of all nodes and the overall situation, therefore obtaining the most suitable feature representation for subsequent recognition. Experiments show that the proposed method is significantly superior to previous state-of-the-art methods in recognition accuracy.