Speech Semantic Communication Based on Swin Transformer
Ziliang Zhou, Shilian Zheng, Jie Chen, Zhijin Zhao, Xiaoniu Yang
Abstract
Semantic communication is a novel communication paradigm which refers to the extraction and encoding of semantic information from the source, and the restoration of information from a semantic perspective at the receiver. In this paper, we propose an end-to-end speech semantic communication system based on Transformer, called DeepSC-TS. It focuses on reconstructing and integrating multi-level information from the transmitted semantic signal at the receiver, effectively eliminating semantic signal noise while preserving the original semantic information. Moreover, it does not increase the data load of the original signal during transmission. Simulation results demonstrate that our proposed DeepSC-TS exhibits outstanding performance in different channel environments and it performs better than an existing speech semantic communication system.