Optimisation of GAN Recovery Network for Semantic Communication

Semantic transmission efficiency becomes a key challenge in the intelligent development of communication systems. In this paper, we propose a GAN recovery network optimisation scheme for semantic communication, designing a multicomponent loss function and bidirectional feature fusion structure to realise the channel adaptive recovery mechanism, and after phased training and parameter optimisation, the system demonstrates excellent semantic recovery capability under poor channel conditions. Experimental verification shows that this scheme significantly improves the semantic fidelity and transmission efficiency, and provides a new idea for the next-generation communication system, which has a broad application prospect.