Scholarlink

<p>Semantic communication enhances robustness by prioritizing the preservation of information semantics, addressing the signal degradation caused by traditional communication systems’ reliance on strict bit-level fidelity under low-signal-to-noise ratio (SNR) environments. Current speech semantic communication systems often face challenges in achieving adequate robustness, particularly under dynamic time-varying channel conditions. This paper introduces the DENOISE framework, designed to enable reliable data transmission in complex, fluctuating channel conditions. DENOISE utilizes a low-complexity, trainable semantic encoder that extracts key speech features, while dynamically suppressing noise through diffusion sampling techniques within a channel-conditioned semantic reconstruction mechanism (CCSRM). This mechanism adapts in real time by leveraging noise parameters based on channel state information (CSI). DENOISE ensures high semantic fidelity and minimizes signal distortion across a wide range of SNRs and channel conditions. Simulation results demonstrate that DENOISE achieves an impressive performance improvement of approximately 3.13&#xa0;dB compared to the DSST system, particularly in low-SNR environments.</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

DENOISE: diffusion-enhanced noise optimization for intelligent semantic-aware communication systems

  • 期刊论文

  • 发表日期：

    2025年9月14日

查看全文
  • Shengliang Wu,
  • Cheng Li,
  • Xin He,
  • Weiwei Jiang,
  • Yujun Zhu,
  • Yong Xu

摘要

Semantic communication enhances robustness by prioritizing the preservation of information semantics, addressing the signal degradation caused by traditional communication systems’ reliance on strict bit-level fidelity under low-signal-to-noise ratio (SNR) environments. Current speech semantic communication systems often face challenges in achieving adequate robustness, particularly under dynamic time-varying channel conditions. This paper introduces the DENOISE framework, designed to enable reliable data transmission in complex, fluctuating channel conditions. DENOISE utilizes a low-complexity, trainable semantic encoder that extracts key speech features, while dynamically suppressing noise through diffusion sampling techniques within a channel-conditioned semantic reconstruction mechanism (CCSRM). This mechanism adapts in real time by leveraging noise parameters based on channel state information (CSI). DENOISE ensures high semantic fidelity and minimizes signal distortion across a wide range of SNRs and channel conditions. Simulation results demonstrate that DENOISE achieves an impressive performance improvement of approximately 3.13 dB compared to the DSST system, particularly in low-SNR environments.