Summary
The 2018 International Symposium on Information Theory and Its Applications　(ISITA2018)
2018
Session Number:Mo-PM-2-1
Session:
Number:Mo-PM-2-1.2
Reliability-Based Parametric LDLC Decoding
Warangrat Wiriya, Brian M. Kurkoski,
pp.188-192
Publication Date:2018/10/18
Online ISSN:2188-5079
DOI:10.34385/proc.55.Mo-PM-2-1.2
PDF download
Summary:
This paper proposes reliability-based parametric decoding of low-density lattice codes (LDLC). We define the reliability of the check-to-variable messages for two purposes. The first one is to choose to approximate the infinite Gaussian mixtures by one or two Gaussians. The reliability of each checkto- variable message is calculated. If there is higher reliability than a fixed threshold value, one Gaussian will be selected; otherwise two Gaussians will be used. The other purpose is for the updating sequence of variable nodes of the parametric shuffled BP (SBP) decoding algorithm. The parametric SBP increases the convergence speed. The updating sequence of SBP follows the order of reliability of the check-to-variable messages from high to low. The numerical results show that the proposed algorithm gives superior performance and lower complexity compared to two or three Gaussian decoding algorithm. At a probability of symbol error equal 10^-4 and n = 100 and 1000, the proposed algorithm gains 0.25 and 0.2 dB, respectively. Moreover, the proposed algorithm provides lower decoding time, fewer number of iterations for convergence and lower memory requirement.