Joint Semantic-Physical Neural Network for Underwater Communication

Semantic coding significantly enhances communication efficiency and reduces data traffic by transmitting essential semantic information. A joint semantic-physical neural network (JSPNN) structure is proposed for autonomous underwater vehicle communications to address two technique issues over the underwater environment, including the coding efficiency for real-time requirements and the environment applicability for transmitting limitations. Both the JSPNN structure and the construction of semantic knowledge base are illustrated. Finally, we indicate some forward issues on this topic.