Two low complexity decoding algorithms for ldpc codes

Abstract

In the present invention, two improved variants of the reliability-based iterative majority-logic decoding algorithm for regular low-density parity-check (LDPC) codes are presented. The new algorithms are obtained by introducing a different reliability measure for each check-sum of the parity-check matrix, and taking it into account in the computation of the extrinsic information that is used to update the reliability measure of each received bit in each iteration. In contrast to the first algorithm, the second algorithm includes check reliability that changes at each iteration. For the tested random and structured LDPC codes, both algorithms, while requiring very little additional computational complexities, achieve a considerable error performance gain over the standard one. More importantly, for short and medium block length LDPC codes of relatively large column weight, both algorithms outperform or perform just as well as the iterative decoding based on belief propagation (IDBP) with less decoding complexity.