有限状态Markov信道上基于噪声软判决的LDPC译码

Decoding of low-density parity-check codes based on noise soft-decision over finite-state Markov channels

针对有限状态Markov信道,提出一种改进的低密度奇偶校验码(low-density parity-check,LDPC)译码算法,并给出其因子图表示.该译码算法包括相互迭代的标准和积译码算法和前向后向算法两个部分.在标准和积译码算法每次迭代后,得到噪声比特的软判决;前向后向算法利用噪声比特的软判决,重新估计发送比特的信道似然比.标准和积译码算法用此重新估计的信道似然比,进行下一次迭代.考虑到因子图中Markov信道节点的引入会引起图中圈个数的增加,进一步提出用基于概率的消息传递策略来更新译码过程中的消息.仿真结果表明,此算法不仅远好于标准的和积译码算法,而且优于采用噪声硬判决的算法.

An improved decoding algorithm for low-density parity check code（LDPC） on finite-state Markov channels was presented and the corresponding factor graph representation was given. It consists of two parts： standard sumproduct algorithm （SPA） for LDPC and forward-backward （FB） algorithm for Markov channel. After each iteration for SPA, FB re-estimated the channel likelihood ratio for the next iteration of SPA based on soft decisions for noise bits from SPA. Considering more cycles in the factor graph with the introduction of the Markov channel node, the probabilistic schedule was used to update the message during the iterations for decoding. Simulation results show that the algorithm not only performs much hetter than the standard sum-product algorithm, but also outperforms the algorithm based on hard decisions for noise bits.