消息复用下的空间耦合LDPC码窗译码优化算法

Improved Window Decoding Algorithm of Spatially Coupled LDPC Code Based on Message Reuse

空间耦合低密度奇偶校验(Spatially Coupled Low Density Parity Check,SC-LDPC)码由于其奇偶校验矩阵呈现对角带非零项的特点,采用窗译码可获得较低的时延及复杂度。针对SC-LDPC码窗译码性能损失、译码复杂度依然较高的问题,提出了消息复用(Message Reuse for Window Decoding,MR-WD)算法和动态多目标符号输出(Dynamic Multi-target-symbol Output for Window Decoding,DMO-WD)算法。消息复用算法以外部对数似然比值作为边信息传入窗口,与传统窗译码相比,译码性能得到了提升,译码复杂度降低。动态多目标符号输出算法在连续若干个窗口均满足奇偶校验方程判定为零时,增加目标符号的输出数量,反之减少目标符号的输出数量。该算法在误码率损失可忽略不计的情况下降低译码复杂度最大达20%,且在动态多目标符号窗译码中使用消息复用更新边信息时,相比于传统窗译码性能提升约0.2 dB,复杂度下降约35%。

Since the parity-check matrix of a spatially coupled low density parity check(SC-LDPC)code is semi-infinite along the diagonal,window decoding can achieve lower complexity and smaller delay.For the problems of performance loss and high decoding complex,two window decoding optimization algorithms,i.e.,message reuse(MR)and dynamic multi-target-symbol output(DMO)for window decoding are proposed.The MR algorithm delivers the edge information in log-likelihood ratio(LLR)of the previous window.Compared with the traditional window decoder,MR improves the decoding performance with lower complexity.In the multi-target-symbol output scenario,MR can save the loss in decoding performance caused by insufficient iterative information.The dynamic multi-target-symbol output(DMO)algorithm increases the number of output target symbols when the parity-check constraint is satisfied for several successive windows,and reduces the number of output target symbols otherwise.The proposed algorithm greatly reduces the decoding complexity with negligible loss in bit error rate.