通信学报 ›› 2023, Vol. 44 ›› Issue (6): 70-76.doi: 10.11959/j.issn.1000-436x.2023106

• 学术论文 • 上一篇    下一篇

知识增强的语义通信接收端设计

李荣鹏1, 汪丙炎1, 张宏纲1,2, 赵志峰1,2   

  1. 1 浙江大学信息与电子工程学院,浙江 杭州 310027
    2 之江实验室,浙江 杭州 311121
  • 修回日期:2023-05-04 出版日期:2023-06-25 发布日期:2023-06-01
  • 作者简介:李荣鹏(1989- ),男,河北衡水人,博士,浙江大学副教授,主要研究方向为智能通信网络、网络智能、网络切片
    汪丙炎(1999- ),男,广东深圳人,浙江大学硕士生,主要研究方向为语义通信、深度学习
    张宏纲(1967- ),男,甘肃兰州人,博士,浙江大学教授,主要研究方向为认知无线电、绿色通信和下一代异构蜂窝网络架构
    赵志峰(1975- ),男,河南洛阳人,博士,之江实验室总工程师,主要研究方向为认知无线电、无线mesh网络和SDN在无线通信中的应用
  • 基金资助:
    国家自然科学基金资助项目(62071425);浙江省“领雁”基金资助项目(2022C01093);华为公司合作基金资助项目;浙江省杰出青年基金资助项目(LR23F010005)

Design of knowledge enhanced semantic communication receiver

Rongpeng LI1, Bingyan WANG1, Honggang ZHANG1,2, Zhifeng ZHAO1,2   

  1. 1 College of Information Science and Electronic Engineering, Zhejiang University, Hangzhou 310027, China
    2 Zhejiang Lab, Hangzhou 311121, China
  • Revised:2023-05-04 Online:2023-06-25 Published:2023-06-01
  • Supported by:
    The National Natural Science Foundation of China(62071425);“Leading Goose” Research and Development Program of Zhejiang Province(2022C01093);Huawei Cooperation Project;Zhejiang Provincial Distinguished Young Scholars Foundation(LR23F010005)

摘要:

针对现有的语义通信系统对先验知识利用不够充分、接收端解码能力有限的问题,提出了一个知识增强的语义通信框架。在这个框架中,接收端可以利用知识库中的先验知识进行语义推理和解码,同时不需要对发送端的神经网络结构进行额外的修改。具体而言,在语义接收端的基础上,设计了一个基于 Transformer 的知识提取器来为接收到的含噪信号寻找语义相关的知识三元组,以用于语义解码。在WebNLG数据集上的仿真结果表明,所提框架在知识图谱增强解码的基础上产生了明显的性能提升。

关键词: 语义通信, 知识图谱, 深度学习, 知识提取, 语义解码

Abstract:

To address the problem that existing semantic communication do not make sufficient use of prior knowledge and have limited decoding capability at the receiver side, a knowledge enhanced semantic communication framework was proposed, in which the receiver could more actively utilize the prior knowledge in the knowledge base for semantic reasoning and decoding, without extra modifications to the neural network structure of the transmitter.Specifically, a transformer-based knowledge extractor was designed to find relevant factual triples for the received noisy signal.Extensive simulation results on the WebNLG dataset demonstrate that the proposed framework has significantly improved performance on the basis of knowledge graph enhanced decoding.

Key words: semantic communication, knowledge graph, deep learning, knowledge extraction, semantic decoding

中图分类号: 

No Suggested Reading articles found!