Online Learning for Wireless Image Semantic Transmission in Vehicular Networks

This paper explores a novel online semantic communication strategy over vehicular networks for robust image transmission over multiple-input multiple-output (MIMO) channels. We incorporate channel state information (CSI), and propose an online feature-and-channel attention (OFCA) network. Both CSI and signal-to-noise ratio (SNR) are treated as side information by the semantic codec to effectively combat MIMO fadings and enhance the reconstructed image quality. Moreover, we propose an online training framework for one-step network weight update, where the online updating is based on real-time given channel samples. Simulation results demonstrate the superiority of the proposed framework over traditional schemes and state-of-the-art deep learning (DL)-based semantic communication frameworks in MIMO fading channels.