Deep Source-Channel Coding for Sentence Semantic Transmission With HARQ

Recently, semantic communication has been brought to the forefront because of its great success in deep learning (DL), especially Transformer. Even if successfully applied sentence transmission reduce errors, existing architecture is usually fixed codeword length and inefficient inflexible for varying length. In this paper, we exploit hybrid automatic repeat request (HARQ) error further. We first combine coding (SC) with Reed Solomon (RS) channel HARQ, called SC-RS-HARQ, which exploits superiority SC reliability conventional methods successfully. Although SC-RS-HARQ easily HARQ systems, also develop an end-to-end architecture, SCHARQ, pursue performance Numerical results demonstrate that SCHARQ significantly reduces required number bits rate. Finally, attempt replace detection from cyclic redundancy check a similarity network Sim32 allow receiver reserve wrong sentences similar information save resources.