数据不均衡条件下数据增强辅助的自动调制识别

打开文本图片集

中图分类号： TN91 文献标志码：A 文章编号： 1000-5013（2026）01-0104-08

Abstract：To address the data imbalance problem in automatic modulation recognition，a data augmentationassisted automatic modulation recognition method is proposed. Based on the Wasserstein generative adversarial network with gradient penalty（WGAN-GP）framework，the generator is first integrated with a self attention mechanism to effectively model the global dependency relationship of signals.Then，a residual feature enhancement module is introduced to solve the gradient vanishing problem using its eficient gradient transfer mechanism. Finally，enhanced data is generated for minority class samples and added to the training dataset to achieve balanced training dataset. The results show that under the condition of imbalanced data，the proposed method achieves superior recognition accuracy.

Keywords：automatic modulation recognition; data imbalance;data augmentation;generative adversarial network

自动调制识别（AMR）是非合作通信场景下的关键技术，其核心任务是通过对接收信号的时频特征分析，准确识别发射端采用的调制制式[1]。（剩余11021字）