Satellite-integrated Internet will enable image transmission for ubiquitous terrestrial users in extreme climates, environments, and terrain in the upcoming 5G advanced (5G-A) and 6G network. However, the challenges p...
详细信息
ISBN:
(纸本)9798350378412
Satellite-integrated Internet will enable image transmission for ubiquitous terrestrial users in extreme climates, environments, and terrain in the upcoming 5G advanced (5G-A) and 6G network. However, the challenges posed by the limited onboard resources and bandwidth of satellite-to-ground link (SGL) in satellite-integrated Internet present significant obstacles for image transmission. To address these challenges, we propose a deep learning-based adaptive semantic coding network (ASCN) for image transmission in satellite-integrated Internet over shadowed-Rician (SR) channel. Our ASCN utilizes a single deep neural network (DNN) to adaptive adjust transmission rate based on input image features and time varying channel state information (CSI) under three types of SR fading levels. Specifically, we design a SR channel ModNet to match the semantic of transmitted image features based on CSI, which can achieve an optimized tradeoff between the image reconstruction quality and network utility. Experimental results demonstrate that our ASCN can effectively learn to match the semantic of transmitted image features based on CSI under three types of SR fading levels, and the image reconstruction quality significantly outperforms the state-of-the-art schemes.
暂无评论