Unequal Error Protection Coding Scheme in Semantic Communication

This paper presents a text semantic communication system that learns the semantic importance of data through a pre-trained neural network and divides semantic information into three layers based on semantic importance. Then unequal error protection (UEP) schemes are constructed using Bose-Chaudhuri-Hocquenghem (BCH) codes of length-31 and length-63. Finally, UEP schemes are applied to the text semantic communication system, and the simulation results indicate that the text average similarity can reach above 92%. Simulation shows that it can improve the transmission reliability of important data, that is, reduce the bit error rate of important data transmission, and further improve the reliability and effectiveness of wireless communication while maintaining complexity.