.png){kind=icon}
.png){kind=icon}
Return
Iterative Neural Rollback Chase–Pyndiah Decoding
.png){kind=icon}
PRE
AI
DOI：10.1109/LCOMM.2025.3638970.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The letter deals with iterative decoding of turbo product codes (TPC). We propose a transformer neural network (NN)-assisted rollback scheme, in which the NN identifies and suppresses extrinsic updates that are likely to introduce decoding errors. Experiments with turbo product code based on (256, 239) extended BCH codes demonstrate that the proposed method improves the bit error rate of Chase-Pyndiah (CP) decoding with $\boldsymbol {p = 6}$ by approximately $\boldsymbol {0.145}$ dB, significantly outperforming conventional CP decoding with $\boldsymbol {p = 7}$ , without modifying the CP decoding pipeline, thereby simplifying integration into existing systems.
Keywords:
Channel decoding
deep neural networks
soft-output decoding
turbo product codes
Journal
IF:
4.4
Papers:
1.3W
Citations:
2.2W