结合CRC校验的LDPC码后处理译码算法

Post-processing decoding algorithm for LDPC codes with CRC checking

为了改善低密度奇偶校验码的错误平层,提出一种结合循环冗余校验位的分层置信传播(cyclic redundancy check layered belief-propagation,简称CRC-LBP)后处理算法。对LBP译码的判决结果执行CRC校验,若校验失败,则对错误帧中变量节点的后验对数似然比值从小到大排序。从排序的节点序列中选取前若干个目标节点组成坏节点集,翻转BN集内信息节点的初始LLR值,并继续译码直至CRC校验成功,打破潜在陷阱集,提高节点可靠度。仿真实验结果表明,在增加少量存储空间的代价下,CRC-LBP算法具有更低的错误平层。

In order to mitigate the error-floor for LDPC codes,a layered belief-propagation post-processing algorithm with cyclic redundancy check was proposed.CRC checking was done at the end of each iteration for LBP decoding,if CRC checking failed,the posteriori logarithm likelihood ratios of variable nodes in the erroneous frames were sorted in ascending orders.Then the several former target nodes from the ordered nodes sequence were selected to form bad nodes set,each initial LLR value for decoded information bits in BN set was flipped and the decoding was done continuously until CRC succeeded,so as to break the potential trapping sets and improve the reliability of variable nodes.The simulation results demonstrate that the CRC-LBP algorithm has lower error-floor at the expense of little extra memory space.