Dynamic Dual Quantization-Domain Decoding for NR LDPC Codes

In this paper, we propose a low-density parity-check (LDPC) decoder with a dynamic dual quantization-domain (DQD) that check nodes are operated in a 3-bit domain while variable nodes are operated in a higher bits domain. The proposed mapping to transform messages from a domain to the other domain gradually changes during the LDPC decoding process and is optimized by differential evolution and machine learning. In addition, for performance improvement, the variable node performs self-correction (SC) that the unreliable messages are erased. Simulation results show that the proposed dynamic DQD-SC decoder has better error correction performance compared to the previously proposed hybrid DQD decoder.