Scholarlink

<p>Automatic Modulation Classification (AMC) is a key component of modern wireless communication systems; however, its performance remains highly sensitive to noise, particularly under low signal-to-noise ratio (SNR) conditions. Most existing approaches rely on benchmark datasets and computationally complex architectures, which limit robustness and practical applicability. To address these challenges, this study proposes an efficient CNN-based AMC framework trained on a customized dataset generated using MATLAB, enabling full control over modulation types, channel conditions, and SNR distributions. STFT-based spectrogram representations are employed to extract discriminative time–frequency features from modulated signals. Experimental results demonstrate that the proposed model achieves a maximum classification accuracy of 99.77% at 15 dB, while maintaining robust performance across an SNR range of -5 to 15 dB, including low-SNR scenarios. Compared with existing methods, the proposed approach improves generalization capability, supports eleven modulation schemes, and reduces dependency on pre-existing datasets, making it more suitable for real-world wireless communication. applications.</p>

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

Powerful deep convolutional neural networks for robust automatic modulation classification using spectrograms

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月17日

查看全文
  • Ola Fadhil Obead,
  • A. M. El-Assy,
  • Hossam El-Din Moustafa,
  • Hala B. Nafea

摘要

Automatic Modulation Classification (AMC) is a key component of modern wireless communication systems; however, its performance remains highly sensitive to noise, particularly under low signal-to-noise ratio (SNR) conditions. Most existing approaches rely on benchmark datasets and computationally complex architectures, which limit robustness and practical applicability. To address these challenges, this study proposes an efficient CNN-based AMC framework trained on a customized dataset generated using MATLAB, enabling full control over modulation types, channel conditions, and SNR distributions. STFT-based spectrogram representations are employed to extract discriminative time–frequency features from modulated signals. Experimental results demonstrate that the proposed model achieves a maximum classification accuracy of 99.77% at 15 dB, while maintaining robust performance across an SNR range of -5 to 15 dB, including low-SNR scenarios. Compared with existing methods, the proposed approach improves generalization capability, supports eleven modulation schemes, and reduces dependency on pre-existing datasets, making it more suitable for real-world wireless communication. applications.