Scholarlink

Automatic Modulation Classification (AMC) is to recognize the modulation type of received signal, which can be used to ensure the wireless security. In this paper, a residual convolutional recurrent neural network (RCRNN) is proposed to make classification, where two cues amplitude/phase (AP) and in-phase/quadrature (IQ) are used together, which has a better classification performance than using only one of them. In addition, a novel shrinkage loss is utilized to optimize distances among different modulations, where the inner products of intra-class are increased with respect to inter-class. Numerical results suggest that our proposed method has made a superior performance.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Multi-loss Learning Based Residual Convolutional Recurrent Neural Network for Automatic Modulation Classification

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Shuo Chang,
  • Sai Huang,
  • Zheng Yang,
  • Weiwei Jiang,
  • Han Zhang,
  • Shaoshuai Fan

摘要

Automatic Modulation Classification (AMC) is to recognize the modulation type of received signal, which can be used to ensure the wireless security. In this paper, a residual convolutional recurrent neural network (RCRNN) is proposed to make classification, where two cues amplitude/phase (AP) and in-phase/quadrature (IQ) are used together, which has a better classification performance than using only one of them. In addition, a novel shrinkage loss is utilized to optimize distances among different modulations, where the inner products of intra-class are increased with respect to inter-class. Numerical results suggest that our proposed method has made a superior performance.