In this paper, we propose a novel joint source-channel coding (JSCC) approach for channel-adaptive digital semantic communications. In semantic communication systems with digital modulation and demodulation, end-to-end training and robust design of JSCC encoder and decoder becomes challenging due to the nonlinearity of modulation and demodulation processes, as well as diverse channel conditions and modulation orders. To address this challenge, we first develop a new demodulation method which assesses the uncertainty of the demodulation output to improve the robustness of the digital semantic communication system. We then devise a robust training strategy that facilitates end-to-end training of the JSCC encoder and decoder, while enhancing their robustness and flexibility. To this end, we model the relationship between the encoder's output and decoder's input using binary symmetric erasure channels and then sample the parameters of these channels from diverse distributions. We also develop a channel-adaptive modulation technique for an inference phase, in order to reduce the communication latency while maintaining task performance. In this technique, we adaptively determine modulation orders for the latent variables based on channel conditions. Using simulations, we demonstrate the superior performance of the proposed JSCC approach for both image classification and reconstruction tasks compared to existing JSCC approaches.

翻译：本文提出了一种面向信道自适应的数字语义通信的新型联合信源信道编码(JSCC)方法。在采用数字调制与解调的语义通信系统中，由于调制解调过程的非线性特性以及信道条件和调制阶数的多样性，JSCC编码器与解码器的端到端训练及鲁棒性设计面临挑战。为解决该问题，我们首先开发了一种新的解调方法，通过评估解调输出的不确定性来提升数字语义通信系统的鲁棒性。随后，我们设计了一种鲁棒训练策略，该策略不仅促进JSCC编码器与解码器的端到端训练，还增强了其鲁棒性与灵活性。为此，我们采用二元对称擦除信道对编码器输出与解码器输入之间的关系进行建模，并从多种分布中采样这些信道的参数。同时，为降低推理阶段的通信时延并保持任务性能，我们提出了一种信道自适应调制技术。该技术根据信道条件自适应确定潜在变量的调制阶数。仿真结果表明，在图像分类与重建任务中，本文提出的JSCC方法相较于现有方法具有更优的性能。