【摘要】 6G语义通信的高效可靠传输需求对于经典分离信源信道定理指导下的传统通信架构形成了挑战，利用端到端优化的JSCC(Joint Source-Channel Coding)能够克服动态信道与资源瓶颈，为6G语义通信提供了新型的解决方案。首先系统梳理了JSCC的最新研究进展，探讨了端到端优化的核心工作机制；然后构建了面向多模态特征的分析框架，并针对文本、语音与图像等典型信源，分别探讨了JSCC的演进范式与典型架构，其中文本场景由字符句法级传输转向语义表示驱动，语音场景结合感知特征面向智能交互与语义理解，图像场景则从像素级重建转向视觉语义保持；最后从模型轻量化、跨模态语义表示、动态环境鲁棒性及算网联合优化等维度，剖析了JSCC面向实际系统落地所面临的关键挑战。研究表明，JSCC方案能够显著提升多模态信息的感知质量与任务效能，并在支撑未来6G语义通信实际部署和演进中具有重要的应用前景与基础性作用。更多还原

【Abstract】 The demand for efficient and reliable transmission in sixth-generation(6G) semantic communications poses challenges to conventional communication architectures guided by the classical source-channel separation theorem. End-to-end optimized joint source-channel coding(JSCC) can overcome dynamic channel conditions and resource constraints, providing a new solution for 6G semantic communications. This paper first systematically reviews recent advances in JSCC and discusses the core working mechanism of end-to-end optimization. A multimodal feature-oriented analytical framework is then constructed, and the evolutionary paradigms and representative architectures of JSCC are discussed for typical sources, including text, speech, and images. In text transmission, JSCC is evolving from character-and syntax-level transmission toward semantic representation-driven transmission. In speech transmission, perceptual features are incorporated to support intelligent interaction and semantic understanding. In image transmission, the focus is shifting from pixel-level reconstruction to visual semantic preservation. Finally, key challenges for practical JSCC deployment are analyzed from the perspectives of model lightweighting, cross-modal semantic representation, robustness in dynamic environments, and joint communication-computation-network optimization. The study shows that JSCC schemes can significantly improve the perceptual quality and task effectiveness of multimodal information, and play a fundamental role in supporting the practical deployment and evolution of future 6G semantic communications.更多还原