Priority-Aware Multi-Layer Transmission in Video Semantic Communications

This paper proposes a priority-aware multi-layer transmission framework to enhance the Quality of Experience (QoE) for video semantic communications over lossy networks. Specifically, the framework employs a dynamic resource allocation strategy that prioritizes video data according to their semantic importance and jointly optimizes redundancy and buffer delay across multiple priority layers. The key contributions are twofold: (1) a novel QoE model that accurately quantifies the differential impacts of semantic data at multiple priority levels on perceptual quality, enabling a more nuanced evaluation of user experience; (2) a cross-layer optimization approach that systematically derives optimal transmission parameters tailored for each semantic priority, thus significantly improving overall system performance. Theoretical analysis demonstrates that the proposed multi-layer framework consistently outperforms conventional single-layer transmission methods. Experimental results in lossy network environments indicate a notable improvement, raising the maximum tolerable packet loss threshold by $12.9 \%$ (from 31% to 35%) and achieving an improved QoE score (from 0.74 to 0.78) under severe packet loss conditions. Additionally, the proposed method visibly enhances the quality of reconstructed video, demonstrating its practical potential for real-world video semantic communication applications.