基于可靠度的LDPC码梯度下降比特翻转译码算法  

Reliability Based on Gradient Descent Bit-flipping Decoding Algorithm for LDPC Codes

作　　者：彭伟夫 张高远[2] 文红[2] 王龙业[2,3] 

机构地区：[1]国家电网四川省电力公司信息通信公司,四川成都610041 [2]电子科技大学通信与抗干扰技术国家重点实验室,四川成都611731 [3]西藏大学工学院,西藏拉萨850000

出　　处：《电视技术》2015年第13期129-134,共6页Video Engineering

基　　金：国家自然科学基金项目(61261021);中央高校基本科研业务费专项(A03008023901004)

摘　　要：仿真结果表明,对于列重和行重较小的低密度奇偶校验(Low Density Parity-check,LDPC)码而言,梯度下降比特翻转(Gradient Descent Bit-flipping,GDBF)译码算法展现出巨大的性能优势,但其对于基于有限域几何构造的列重和行重较大的LDPC码则性能损失严重。该文首先分析指出,对于大列重LDPC码而言,翻转函数中的"互相关项"和"双极性校验子求和项"之间的"不匹配"是造成性能损失的主要原因。其次,引入一种可靠性度量对双极性校验子进行加权,上述"不匹配"现象得到有效削弱,从而改善GDBF算法对大列重LDPC码的译码性能。仿真结果表明,在加性高斯白噪声信道下,相比于传统的GDBF算法,新提出的算法在误比特率为10^(-5)时可获得0.8 dB的增益。Simulation results show that the gradient descent bit-flipping （GDBF）decoding algorithm performs extraordinarily well when it is used for some low density parity-check （LDPC） codes with low row/colum weight. However, the performace will degrade when GDBF decoding is used for large row/colum weight finite-geometry （FG） LDPC codes. The mismatch between ＂ the correlation term＂ and ＂the sum term of bipolar syndromes＂ in the inversion function of GDBF algorithm for large row/colum LDPC codes is first theoretically analyzed in this paper, which is supposed to be the largest contribution to the performance degradation. Secondly, a reliability metric is considered to weight the bipolar syndrome, and improvement in performance is observed, which benefits from sufficiently impairing the aforementioned mismatch. Simulations show that the performance improvement is about 0. 8 dB at BER of 10^-5 over an AWGN channel.

关 键 词：LDPC码 梯度下降 不匹配 加权梯度下降 比特翻转 

分 类 号：TN911.22[电子电信—通信与信息系统]