基于AI辅助的可扩展视频语义通信系统  

AI-Empowered Scalable Video Semantic Communication Systems

作　　者：韩雪 吴泳澎 施雨轩 张文军[1] HAN Xue;WU Yongpeng;SHI Yuxuan;ZHANG Wenjun(School of Electronic Information and Electrical Engineering,Shanghai 200240,China)

机构地区：[1]上海交通大学电子信息与电气工程学院,上海200240

出　　处：《移动通信》2023年第6期27-34,共8页Mobile Communications

基　　金：国家重点研发计划(2018YFB1801102);国家自然科学基金项目(62122052,62071289);中央高校基本科研业务费专项资金项目,高等学校学科创新引智计划(BP0719010);上海市自然科学基金项目(22DZ2229005)。

摘　　要：语义通信作为一种面向多智能体交互的新型通信范式,对未来通信实现高效传输提供了新的解决方法。由此,提出了一个AI辅助的可扩展视频语义通信系统。该系统能够基于现有的信源信道分离编码方案(如传统的高效视频编码HEVC与低密度奇偶校验编码LDPC),与AI辅助的可学习语义编解码器相结合,实现视频流与语义流的共同传输。其中可学习的语义数据流是视频帧中针对特定任务提取的语义信息,能够较好地提升接收端精确重建与下游任务的性能。实验证明,该视频语义通信框架相较于传统通信,在重建任务性能相差无几的情况下具有更好的下游任务表现;而相较于现有的端到端的基于深度学习的语义通信系统,性能相近的同时能够基于多种传统视频编解码标准进行扩展设计,具有与传统压缩方案更强的耦合性与适应更多场景的能力。As a new communication paradigm for multi-agent interaction,semantic communication offers a promising solution for efficient transmission in future communications.In this paper,we propose an Al-assisted scalable video semantic communication system.Based on existing source-channel separation coding schemes,like traditional high-efficiency video coding(HEVC)and low-density parity-check coding(LDPC),the proposed system can be combined with AI-assisted learnable semantic encoders and decoders to realize the joint transmission of video streams and semantic streams.Among them,learnable semantic data streams are the sematic information extracted from video frames for specific tasks,thereby improving the reconstruction accuracy and downstream task performance at the receiver.Experimental results show that our proposed system outperforms traditional communication frameworks in terms of downstream task performance while maintaining similar reconstruction performance.Compared with the existing end-to-end deep learning-based semantic communication systems,our approach can be extended and designed based on various traditional video coding standards and has stronger coupling and adaptability to more scenarios while presenting the similar performance.

关 键 词：语义通信 视频压缩 深度学习 无线通信 

分 类 号：TN915[电子电信—通信与信息系统]