摘要
联合信源信道编码作为语义通信的关键研究方向,已取得初步研究成果。然而,随着图像分辨率的提升,传统的基于卷积神经网络的联合信源信道编码算法在图像语义特征提取方面表现出局限性。为了解决该问题,文章提出了一种基于Swin Transformer的联合信源信道编码算法。该算法首先利用多尺度大核注意力机制初步捕获图像的局部信息和长距离依赖性,然后通过Swin Transformer进一步对图像语义特征进行分层提取和自适应码率编码。实验结果表明,在AWGN和Rayleigh信道模型下,所提出的算法在PSNR和MS-SSIM指标上均优于传统的算法。
Joint Source-Channel Coding(JSCC),as a key research direction in semantic communication,has achieved preliminary research results.However,with the increasing resolution of images,traditional JSCC algorithms based on Convolutional Neural Network(CNN)exhibit limitations in extracting image semantic features.To address this issue,this paper proposes a JSCC algorithm based on Swin Transformer.The algorithm firstly utilizes a Multi-Scale Large Kernel Attention(MLKA)mechanism to initially capture the local information and long-range dependencies of images.Subsequently,Swin Transformer is employed to further hierarchically extract image semantic features and perform adaptive rate coding.Experimental results demonstrate that,under the channel models of Additive White Gaussian Noise(AWGN)and Rayleigh,the proposed algorithm outperforms traditional algorithms in terms of Peak Signal-to-Noise Ratio(PSNR)and Multi-Scale Structural Similarity Index Measure(MS-SSIM).
作者
廖潇
李智
LIAO Xiao;LI Zhi(College of Electronics and Information Engineering,Sichuan University,Chengdu 610065,China)
出处
《现代信息科技》
2025年第7期1-4,共4页
Modern Information Technology
