高效融合全局和局部上下文特征的语义通信系统

Semantic communication system with efficient integration of global and local context features

采用基于深度学习的端到端一体化设计方法,提出了一种基于扩展上下文语义特征的通信系统。区别于已有研究中只关注局部上下文而忽视全局上下文的问题,所提系统有机地融合了局部和全局上下文知识,利用扩展上下文知识进行语义编码和语义译码,提升了语义通信系统的可靠性。在发送端,采用基于扩展上下文的语义编码,实现高效的语义表征。在接收端,结合历史通信文本挖掘机制、上下文语义特征学习机制和基于启发式图搜索的译码策略,提升语义推理的准确性。与现有语义通信系统仿真对比结果表明,相较于传统通信系统和现有语义通信系统,所提系统在低信噪比下显著提升了通信系统的可靠性。

A communication system based on extended contextual semantic features was proposed by using an end-to-end integrated design method based on deep learning.Unlike existing research that focused only on local context while neglecting global context,the proposed system integrated both local and global contextual knowledge,semantic encoding and decoding was utilized by extended contextual knowledge,thereby enhancing the reliability of the semantic communication system.At the transmitter,efficient semantic representation was achieved through extended contextual semantic encoding.At the receiver,the accuracy of semantic inference was improved by combining mechanisms such as historical communication text mining,contextual semantic feature learning,and heuristic graph-based decoding strategy.When comparing with the traditional communication system and the existing semantic communication systems,simulation results demonstrate that the proposed system significantly improves the reliability of the communication system under low signal-to-noise ratio.