Dynamic check message majority-logic decoding algorithm for non-binary LDPC codes

Yichao Lu, Xiao Peng, Guifen Tian, Satoshi Goto

Research output: Contribution to journalArticlepeer-review


Majority-logic algorithms are devised for decoding nonbinary LDPC codes in order to reduce computational complexity. However, compared with conventional belief propagation algorithms, majority-logic algorithms suffer from severe bit error performance degradation. This paper presents a low-complexity reliability-based algorithm aiming at improving error correcting ability of majority-logic algorithms. Reliability measures for check nodes are novelly introduced to realize mutual update between variable message and check message, and hence more efficient reliability propagation can be achieved, similar to belief-propagation algorithm. Simulation results on NB-LDPC codes with different characteristics demonstrate that our algorithm can reduce the bit error ratio by more than one order of magnitude and the coding gain enhancement over ISRB-MLGD can reach 0.2-2.0 dB, compared with both the ISRB-MLGD and IISRB-MLGD algorithms. Moreover, simulations on typical LDPC codes show that the computational complexity of the proposed algorithm is closely equivalent to ISRB-MLGD algorithm, and is less than 10% of Min-max algorithm. As a result, the proposed algorithm achieves a more efficient trade-off between decoding computational complexity and error performance.

Original languageEnglish
Pages (from-to)1356-1364
Number of pages9
JournalIEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences
Issue number6
Publication statusPublished - 2014


  • Iterative decoding
  • Majority-logic decoding
  • Message-passing algorithms
  • Non-binary LDPC codes

ASJC Scopus subject areas

  • Electrical and Electronic Engineering
  • Computer Graphics and Computer-Aided Design
  • Applied Mathematics
  • Signal Processing


Dive into the research topics of 'Dynamic check message majority-logic decoding algorithm for non-binary LDPC codes'. Together they form a unique fingerprint.

Cite this