An Improved Reliability-Based Decoding Algorithm for NB-LDPC Codes

When non-binary low-density parity-check (NB-LDPC) codes with low column weights are adopted, the iterative reliability-based majority-logic decoding (IRB-MLGD) algorithms suffer from severe performance degradation compared to message passing algorithms. In this brief, based on the improved iterative soft reliability-based (IISRB)-MLGD algorithm, we propose a perturbation-injected (P)-IISRB algorithm, where a regular perturbation is introduced to alleviate the periodic point problem that is the main reason for IISRB decoding failure. Compared with the conventional IISRB, the new algorithm significantly improves the decoding performance, lowering the error-floor by at least two orders of magnitude for the example code. Besides, instead of traditionally pre-computing and storing all the q-ary channel reliability measures, a real-time computing method is presented, which reduces the memory cost from
O
( q) to
O
( log
₂
q).