Scholarlink

<p>Cognitive radio systems require efficient spectrum sensing and signal reconstruction to enable dynamic spectrum access and accurate detection of primary user activity. Conventional compressed sensing methods reduce sampling rates but suffer from high computational comp Cognitive radio systems require efficient spectrum sensing and signal reconstruction to enable dynamic spectrum access and reliable detection of primary user activity. Conventional compressed sensing techniques reduce sampling rates but often rely on computationally intensive iterative optimization, which limits their suitability for real-time deployment. To address this limitation, this paper proposes a deep learning–driven framework based on a convolutional neural network (CNN) that directly reconstructs signals from compressed measurements in a non-iterative manner. The proposed model learns an end-to-end mapping between compressed observations and original signals, enabling fast signal recovery followed by spectrum sensing. Simulation results demonstrate improved reconstruction accuracy and detection performance compared to selected CNN-based baselines, including AlexNet, VGG-16, and ResNet-50, achieving a mean squared error (MSE) of 0.0135 and a detection probability of 0.99 at 10&#xa0;dB SNR, with a false alarm rate of 0.05. While the results indicate the potential of the proposed approach for low-latency spectrum sensing in cognitive radio environments, further evaluation against model-driven spectrum sensing methods and hardware-level validation remain important directions for future work.</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

Deep learning-driven compressed sensing using U-Net for cognitive radio spectrum detection

  • 期刊论文

  • 发表日期：

    2026年2月3日

查看全文
  • Kavitha Veerappan,
  • G. Seetharaman

摘要

Cognitive radio systems require efficient spectrum sensing and signal reconstruction to enable dynamic spectrum access and accurate detection of primary user activity. Conventional compressed sensing methods reduce sampling rates but suffer from high computational comp Cognitive radio systems require efficient spectrum sensing and signal reconstruction to enable dynamic spectrum access and reliable detection of primary user activity. Conventional compressed sensing techniques reduce sampling rates but often rely on computationally intensive iterative optimization, which limits their suitability for real-time deployment. To address this limitation, this paper proposes a deep learning–driven framework based on a convolutional neural network (CNN) that directly reconstructs signals from compressed measurements in a non-iterative manner. The proposed model learns an end-to-end mapping between compressed observations and original signals, enabling fast signal recovery followed by spectrum sensing. Simulation results demonstrate improved reconstruction accuracy and detection performance compared to selected CNN-based baselines, including AlexNet, VGG-16, and ResNet-50, achieving a mean squared error (MSE) of 0.0135 and a detection probability of 0.99 at 10 dB SNR, with a false alarm rate of 0.05. While the results indicate the potential of the proposed approach for low-latency spectrum sensing in cognitive radio environments, further evaluation against model-driven spectrum sensing methods and hardware-level validation remain important directions for future work.