Semantic Communication System for 3D Video

Semantic communication systems are targeted to be the new frontier of 6G communications. They will solve the issue that the current 5G systems are getting close to the Shannon limit while the amount of data is constantly growing. The semantic representation of the world will allow understanding the data by the end nodes as well as the network itself. Using common knowledge of the world by all network nodes will result in data compression beyond the existing compression types. This will be achieved by filtering redundant and irrelevant information and extracting the meaning. Undoubtedly, the main force behind the semantic communications is the development of AI. It can understand the world and because of this, it is in the heart of this new paradigm, semantic communications. Considering the emerging communication applications, such as the metaverse, compression and transmission of 3D data is expected to be one of the primary types of data in 6G. This type of data has properties that are not inherent to traditional 2D images and video, and therefore necessitates the use of new compression types. In this paper, we propose a semantic representation-based system for semantic communication of 3D video data. For encoding and decoding 3D data in the system, graph-convolutional neural networks are used. A dataset of 3D models of moving human bodies that were artificially created is utilized to test it.