基于SwinTransformer的联合信源信道编码算法

打开文本图片集

中图分类号：TN911.22；TP391.4 文献标识码：A 文章编号：2096-4706（2025）07-0001-04

Abstract： JointSource-Channel Coding （JSCC），asa keyresearch 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）exhibitlimitations inextractingimagesemanticfeatures.Toadressthisissue，this paper proposes a JSCCalgorithm based on Swin Transformer.The algorithm firstlyutilizesa Multi-Scale Large Kemel Attention（MLKA）mechanismtoinitiallcapture thelocal informationand long-rangedependenciesofimages.Subsequently SwinTransformer is employed to further hierarchically extract image semantic features and perform adaptiverate coding. Experimentalresultsdemonstrate that，underthechannelmodelsofAditiveWhite GausianNoise （AWGN）andRayleighthe proposedalgorithmoutperformstraditionalalgorithmsin termsofPeak Signal-to-NoiseRatio （PSNR）andMulti-Scale Structural Similarity Index Measure （MS-SSIM）.

Keywords： Joint Source-Channel Coding; Swin Transformer; Multi-Scale Large-Kernel Attention

0 引言

随着信息技术的飞速发展，通信系统的性能要求日益提高。（剩余6464字）