LDPC码的一种低复杂度归一化最小和译码算法  

A low-complexity normalized min-sum decoding algorithm for LDPC codes

作　　者：陈发堂[1] 刘一帆 唐成 CHEN Fatang;LIU Yifan;TANG Cheng(School of Communication and Information Engineering,Chongqing University of Posts and Telecommunications,Chongqing 400065,P.R.China)

机构地区：[1]重庆邮电大学通信与信息工程学院

出　　处：《重庆邮电大学学报（自然科学版）》2020年第1期92-98,共7页Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition)

基　　金：国家科技重大专项(2012ZX03001024)~~

摘　　要：为了满足5G新无线对标准低密度奇偶校验(low-density parity-check,LDPC)码纠错译码器的要求,提出一种基于归一化最小和算法的单最小值算法。利用一次绝对最小值计算和近似第二最小值代替两次最小值计算,减少译码器的运算复杂度。通过密度进化理论计算归一化因子α,利用加权平均修正出最优的α值提前存储,可以在不消耗额外计算资源的前提下改善由于使用单最小值而损失的性能。提出一种分层译码器结构,利用值重用技术实现减少内存和计算资源消耗。仿真结果证明,在比特错误率(bit error ratio,BER)为10-5时,所提算法比现有的单最小值算法有大约0.2 dB的增益,也比传统归一化最小和算法拥有更好的译码性能和收敛速度。This paper proposes a single minimum algorithm based on normalized min-sum algorithm for decoding low-density parity check codes,which meet the requirement of error correction decoder for 5G new radio.The proposed algorithm uses the first absolute minimum value and the second approximate minimum value to replace the calculation of two minimum values.This method can dramatically reduce complexity.In addition,the normalization factorαis calculated by density evolution theory,and then the weighted average is used to optimizeαvalue and stored in advance.This algorithm improves the performance of single minimum algorithm without consuming additional computing resources.In addition,we propose a layered decoding structure,which can reduce memory and complexity of implementation by using reuse message value.The simulation results show that the proposed algorithm has a gain of about 0.2 dB compared with the current single minimum algorithm when the BER is 10-5.Especially,the proposed algorithm has better error correction performance and convergence speed than the traditional normalized min-sum algorithm.

关 键 词：归一化最小和算法 密度进化 低复杂度 

分 类 号：TN929.5[电子电信—通信与信息系统]