Multi-Modal Semantic Communication With Point Cloud Diffusion

Weiyan Feng; Zhenyu Shi

ICASSP 2025accepted0 citations

Multi-Modal Semantic Communication With Point Cloud Diffusion

Abstract

Semantic communication optimizes efficiency by selectively conveying the essential elements of information, thereby eliminating superfluous data transmission and enhancing the overall effectiveness of the communication process. The surge in demand for high-precision perception within the AR/VR realm is fueling an escalating need for the transmission of 3D point cloud data. In this study, we propose a deep neural network integrating image and point cloud semantic information to reconstruct point cloud based on diffusion model. Our network has achieved a remarkable reduction in chamfer distance ranging from 35.7% to 71.2% across various signal-to-noise ratio conditions. Code available at 2730938958/DiffusionPcd.

BibTeX

@inproceedings{icassp2025_multimodalsemant,
  title = {Multi-Modal Semantic Communication With Point Cloud Diffusion},
  author = {Weiyan Feng and Zhenyu Shi},
  booktitle = {ICASSP 2025},
  year = {2025}
}