NOMA Enhanced Semantic Communication with Swin Transformer in STINs

Satellite-terrestrial integrated networks (STINs), as a type of non-terrestrial networks (NTN), can provide seamless global communication services. These networks are envisioned as a crucial component of future 6G systems. However, STINs face significant challenges, such as Doppler frequency shifts and other interferences, which can degrade communication performance. Additionally, the exponential growth of internet of things smart devices has intensified the scarcity of spectrum resources, a finite and non-renewable asset, posing a bottleneck for 6G development. To address these challenges, this paper proposes a non-orthogonal multiple access (NOMA) enhanced semantic communication architecture in STINs (SeC-NOMA). Furthermore, given the high model complexity and poor reconstruction quality of existing image semantic communication models, we propose a Swin Transformer-based semantic communication model in STINs (SwinSeC). The simulation results demonstrate that the SwinSeC model achieves at least a 12 dB improvement in PSNR and a 60% improvement in SSIM compared to traditional models, and delivers superior transmission performance under various channel conditions. In addition, NOMA technology can significantly improve the spectrum utilization of SwinSeC.