Modulation recognition of underwater acoustic communication signals based on deep learning

In recent years, research on modulation signal recognition using deep learning (DL) has achieved significant success. However, the complex environment of underwater channels presents substantial challenges for modulation recognition. To address this issue, this paper proposes a multi-scale feature fusion hybrid model (HM) based on Gram angle field, Markov transition field, and recurrence plot (GMR). The hybrid model is used to fuse time domain and frequency domain features and integrate multi-scale features into low-dimensional features. Results from lake trials in Qiandao Lake demonstrate that the network achieves a recognition accuracy of 94.31% under real underwater acoustic channel conditions. Comparative experiments with other methods further prove the superiority of the proposed method across multiple evaluation metrics.