摘要
作为一种新兴的通信范式,语义通信利用深度学习模型实现联合信源信道编码(Joint Source-Channel Coding,JSCC)。JSCC在无线图像传输中表现出优异的压缩和抗干扰能力,尤其在低信噪比(Signal to Noise Ratio,SNR)下。为提升对高分辨率图像语义特征的提取能力,在基于Swin Transformer的JSCC架构上添加了残差连接,提出了一种新的JSCC架构——Swinformer-R,设计了仿真实验。与3种基准方案对比结果表明,所提方案的峰值SNR(Peak SNR Ratio,PSNR)和多尺度结构相似性指数(Multi-Scale Structural Similarity Index Measure,MS-SSIM)在不同传输信道、不同SNR、不同分辨率图像传输上表现最佳,Swinformer-R架构在提升图像重建质量方面具有显著的潜力和优势。
As an emerging communication paradigm,semantic communication utilizes deep learning models for Joint Source-Channel Coding(JSCC).JSCC has shown excellent compression and interference resistance in wireless image transmission,especially in low Signal to Noise Ratio(SNR)environments.To enhance the ability to extract semantic features from high-resolution images,residual connections are added to a JSCC architecture based on Swin Transformer,a new JSCC architecture called Swinformer-R is proposed,and simulation experiments are designed.Results of comparison with three benchmark schemes demonstrate that the proposed method achieves the best Peak SNR(PSNR)and Multi-Scale Structural Similarity Index(MS-SSIM)across different transmission channels,SNR,and image resolutions.Therefore,the Swinformer-R architecture has significant potential and advantages in improving image reconstruction quality.
作者
香晏
赵响
黄军韬
XIANG Yan;ZHAO Xiang;HUANG Juntao(Guangxi Key Laboratory of Precision Navigation Technology and Application,Guilin University of Electronic Technology,Guilin 541004,China)
出处
《无线电工程》
2025年第5期905-912,共8页
Radio Engineering
基金
国家自然科学基金地区科学基金项目(61961007)。
