面向6G智能内生的隐性语义认知通信  

Implicit Semantic-aware Communication for 6G AI-Native Networks

作　　者：孙子剑 廖逸玮 鲁智敏 肖泳 石光明 高大化 SUN Zijian;LIAO Yiwei;LU Zhimin;XIAO Yong;SHI Guangming;GAO Dahua(Peng Cheng Laboratory,Shenzhen 518055,China;School of Electronic Information and Communications,Huazhong University of Science and Technology,Wuhan 430074,China;School of AI,Xidian University,Xi’an 710071,China;Pazhou Laboratory(Huangpu),Guangzhou 510335,China)

机构地区：[1]鹏城实验室,广东深圳518055 [2]华中科技大学电信学院,湖北武汉430074 [3]西安电子科技大学人工智能学院,陕西西安710071 [4]琶洲实验室(黄埔),广东广州510335

出　　处：《移动通信》2023年第4期7-13,共7页Mobile Communications

基　　金：国家自然科学基金“面向6G群体智能资源共享博弈基础理论研究”(62071193);国家自然科学基金“面向智能语义理解的计算成像方法研究”(61976169);国家自然科学基金“语义通信基础理论与方法研究”(62293483);国家自然科学基金“基于多通道压缩感知的高分辨高动态范围红外成像方法研究”(61871304);鹏城实验室重大攻关项目(PCL2021A12)。

摘　　要：语义通信的目标在于识别信息蕴藏的含义并传递给目标用户。这种基于语义理解的新型通信范式可以实现信息的高效传输、精准表达与智能处理,进而使得网络本身能够理解用户的意图和需求、提供更加个性化和智能化服务,符合人们对未来智能内生通信网络的需求。在目前语义通信领域中,大部分工作着重于传输直接从源信号中识别的显性语义(如标签以及信号特征)。针对隐性语义通信问题展开研究,提出了一种面向智能内生的隐性语义认知通信架构,即通过将信源信号无法识别的隐性关系与其关联的语义信息传递给目标用户,实现了隐性语义的传输与恢复。与传统通信致力于追求通信总量最大化不同,本架构致力于帮助接收端学习一个基于有限信息自动生成隐性语义推理规则。在此架构下,接收端用户始终可以学习与信源用户相匹配的推理规则。实验结果表明,所提出的架构在目标用户的符号误码率方面取得了较大改进,优于现有的非推理通信解决方案。The goal of semantic communication is to identify and deliver the meaning of information to the target user.This new communication paradigm based on semantic understanding enables efficient transmission,precise expression and intelligent processing of information.As a result,the network itself can understand the user’s intentions and needs and provide more personalized and intelligent services,which meets people’s requirements for a future AI-native communication network.In the current field of semantic communication,most work focuses on transmitting explicit semantics that can be directly identified from source signals,e.g.,labels and signal features.This paper focuses on implicit semantic communication and proposes an implicit semantic-aware architecture for 6G AI-native networks.Namely,by transmitting the implicit relationship that cannot be identified from source signals and its associated semantic information to target users,the transmission and recovery of implicit semantics is realized.Unlike traditional communication which strives to maximize total communication volume,this architecture aims to help the target user learn an inference rule for generating implicit semantics based on limited information.Through this architecture,the receiver user can always learn an inference rule that matches the source user.Experimental results show the proposed architecture has achieved significant improvement in symbol error rate for target users,which outperforming the existing non-inference-based communication solutions.

关 键 词：智能内生 语义通信 隐性语义认知通信 

分 类 号：TN92[电子电信—通信与信息系统]