鲁棒语义传输:跨域协作的联合源信道编码OA

[Objective]Realizing high quality and highly reliable information transmission is an important goal in the field of se-mantic communication.Deep Joint Source Channel Coding(DeepJSCC)has emerged as an effective method for semantic com-munication and has made significant progress.However,existing DeepJSCC-based semantic communication methods still face the problem of semantic distortion caused by channel interference in low Signal-to-Noise Ratio(SNR)environments,making it difficult to achieve the desired quality of semantic transmission,thereby limiting the reliability and accuracy of communication.To address this issue,this paper aims to design a novel DeepJSCC framework that effectively suppresses the interference of channel noise on semantic information,improving the robustness of semantic communication systems.[Methods]The proposed Deep-JSCC framework integrates both spatial and frequency domain perspectives,enabling comprehensive and efficient representation and transmission of semantic information.Specifically,in the spatial domain,the framework efficiently extracts global and local semantic features from images,ensuring that semantic information is fully preserved during the encoding stage.In the frequency domain,it precisely identifies the frequency components,enabling accurate discrimination of the frequency components that have the most significant impact on the decoding task.Consequently,it enhances the expression of core semantic frequency components while suppressing the noise frequencies,significantly reducing the semantic distortion caused by channel noise.[Results]We evaluated the performance of the proposed method on public datasets and compared it with existing advanced semantic communica-tion methods.The experimental results demonstrate that,compared to existing DeepJSCC methods,the proposed framework can significantly improve the accuracy of semantic information transmission in adverse communication environments(such as low SNR),effectively mitigating the impact of semantic distortion on communication quality,thereby increasing the robustness of se-mantic communication systems.[Conclusion]The proposed DeepJSCC framework integrates the advantages of both spatial and frequency domains.Through an innovative coding strategy,it achieves efficient semantic feature extraction and enhancement of core semantic frequency components,greatly improving the robustness of semantic communication in adverse environments.This method complements existing DeepJSCC methods rather than replacing them,providing a new solution for the reliability and high-quality transmission of semantic communication systems.Our work provides a new solution for the reliability and high-quali-ty transmission of semantic communication systems.