基于GRU网络辅助和路径度量交换的极化码SCLF算法

SCLF algorithm for polar code based on GRU network assistance and path metric exchange

为了解决极化码的现有SCLF算法传输性能较低,翻转集构建不精确,复杂度过高的问题,提出了一种基于GRU网络辅助和路径度量交换串行抵消列表比特翻转(PME-SCLF)算法。首先,重新划分译码状态,将路径度量交换和反向判决的思想结合,提出了一种位翻转策略。其次,通过训练GRU网络定位译码中的第一个错误位,提出了一种翻转集构造方法。最后,对每次低阶翻转的路径度量进行排序,固定低阶可靠位后再进行高阶翻转,提出了一种多比特翻转规则。仿真结果表明,与现有SCLF算法相比,在低信噪比下,所提算法识别第一个错误比特的准确率提高了18~24%,在单比特和多比特翻转下,所提算法分别有最高0.3 dB和0.2 dB的性能增益,且在线译码复杂度较低。

In order to solve the problems of low transmission performance,inaccurate flip set construction and high complexity of existing polar codes successive cancellation list bit-flip(SCLF)algorithms,a SCLF algorithm for polar codes based on GRU network assistance and path metric exchange was proposed.First the decoding state was redivided,and a bit-flip strategy was proposed by combining the ideas of path metric exchange and reverse judgment.Secondly,a flip set construction method was proposed by training the GRU network to locate the first erroneous bit in the decoding.Finally,a multi-bit flipping rule was proposed by sorting the path metric for each low-order flip and fixing the low-order reliable bits before performing the high-order flip.Simulation results show that compared with the existing SCLF algorithms,the proposed algorithm improves the accuracy of identifying the first error bit by 18~24%at low signal noise ratio.Under single-bit and multi-bit flipping,the proposed algorithm has a performance gain of up to 0.3 dB and 0.2 dB,respectively,and the online decoding complexity is lower.