Automatic modulation classification method combining multimodal information and deep learning

In this paper, we propose a data-driven automatic modulation classification (AMC) algorithm for cross-domain modulation classification scenarios that involve variations in signal symbol rate between the source domain and the target domain. In real-world situations, disparities in signal transmission, differences in transmitter and receiver equipment, and interference factors often lead to shifts in data distribution, creating distinct domains. Existing deep learning (DL)-based intelligent modulation classification algorithms tend to experience performance decline or even fail in such cross-domain scenarios. Therefore, we focus on data preprocessing and extract differential feature sequences of signals as network inputs, guiding the model to learn the intrinsic characteristics of signals in challenging scenarios and developing a more robust AMC algorithm. Experimental results demonstrate that our approach not only achieves better accuracy and robustness in cross-domain scenarios involving symbol rate variations but also has a simple network structure.