【摘要】 语义通信（SC, Semantic Communication）是一项旨在突破香农极限的前沿技术。传统的SC策略通常侧重于减少原始数据与重建数据之间的信号失真，而忽略感知质量。为解决这一问题，提出了一种基于人工智能生成内容（AIGC, Artificial Intelligence Generative Content）技术的图像语义通信系统，该系统采用经典的DeepJSCC网络结构，有效降低通信资源消耗，并将扩散模型（DM, Diffusion Model）部署在解码器中，利用其强大的分布映射能力估计一个紧凑的条件向量指导解码端的图像恢复，从而合成高保真的交付图像。此方法提高了数据传输的效率，优化了从有损信号中恢复复杂语义内容的能力，从而在资源有限的无线网络中提供更智能的通信服务。实验结果表明，与现有方法相比，提出的ISC-G在高分辨率的数据集中表现出优越的传输有效性和通信内容质量。更多还原

【Abstract】 Semantic communication (SC) is a cutting-edge technology aimed at breaking the Shannon limit.Traditional SC strategies typically focus on reducing signal distortion between the original and reconstructed data while often neglecting perceptual quality.To address this issue,this paper proposes an image semantic communication system based on artificial intelligence generative content (AIGC) technology.The proposed system utilizes the classical DeepJSCC network structure to reduce communication resource consumption effectively and deploys the diffusion model (DM) in the decoder to synthesize high-fidelity delivered images by using its powerful distribution mapping capability to estimate a compact conditional vector to guide the image recovery at the decoding end.This approach enhances data transmission efficiency and optimizes the recovery of complex semantic content from lossy signals,providing smarter communication services in resource-constrained wireless networks.Experimental results show that the proposed ISC-G exhibits superior transmission effectiveness and communication content quality in high-resolution datasets compared to existing methods.更多还原