融合知识图谱的语义通信系统

Semantic Communication System Incorporating Knowledge Graphs

知识的理解与处理是语义通信研究的关键问题之一。针对现有语义通信系统在知识表示和知识处理方面的不足,提出了一种融合知识图谱的语义通信框架。在这一框架中,定义了统一语义表示空间的概念,知识库中的实体和接收信号被映射到该空间中,并通过基于对比学习的训练过程以建立关联。在通信过程中,接收端会根据接收到的信号从知识库中检索与之相关的实体,随后通过推理和预测来揭示这些实体之间的关系,从而为解码过程提供有效的辅助信息,同时实时更新知识库。此外,还探讨了基于大语言模型进行数据增强的方案。最后,进行了仿真实验,相较于未经知识增强的语义通信系统,所提系统在较低信噪比下能够取得5%左右的性能提升。结果表明所提系统能够有效地提取相关知识,从而增强语义通信系统接收端的解码能力。

The understanding and processing of knowledge is one of the pivotal concerns in semantic communication research.Aiming at the shortcomings of existing semantic communication systems in knowledge representation and processing,a semantic communication framework incorporating knowledge graph is proposed.In this framework,the concept of a unified semantic representation space is defined,where the entities from the knowledge base and received signals are mapped,and the associations are established through a training process based on contrastive learning.During the communication process,the receiver retrieves the entities associated with the received signals,and subsequently the relationships among these entities are revealed through inference and prediction,thereby providing effective auxiliary information for the decoding process while dynamically updating the knowledge base.Additionally,a solution involving data augmentation using large language models is explored.Finally,simulations affirm that the proposed system approximately attains a performance improvement of 5%under lower signal-to-noise ratios in comparison with the systems lacking knowledge enhancement.The results indicate that the proposed system can effectively extract relevant knowledge,thereby enhancing the decoding capability of the semantic communication system at the receiver side.