Joint Semantic-Channel Coding and Modulation for Point Cloud

Despite remarkable progress in semantic communication, most current deep learning-enabled semantic communication approaches map source information into continuous channel symbols, posing significant challenges for hardware implementation and integration with existing digital communication systems. In this article, we tackle this issue with a joint semantic-channel coding and modulation scheme, enabling the transmitter to generate digital constellation points. Furthermore, a rate allocation model is proposed for assigning different lengths of constellation point sequences to source information with varying semantics. Moreover, motivated by traditional adaptive modulation schemes, we elaborated a channel adaption model to match the modulation with channel conditions. To verify our designs, point cloud, a prevailing three-dimensional representation, which urgently requires a new transmission method due to its large data volume, is investigated. Building upon the proposed methods, a semantic communication system for point cloud is developed. Simulation results demonstrate that the system outperforms existing point cloud transmission methods in terms of efficiency and reliability.