摘要
现有的低密度奇偶校验码(Low Density Parity Check,LDPC)闭集识别方法中,通过直接采用后验概率构建模型计算结果累加的方式会导致在较小码字数量时识别性能较差,通过中心极限定理分布拟合的参数分析方法在部分条件下偏离实际分布,从而无法获得更高的识别性能。基于这些问题,文中提出了一种基于分布参数的LDPC闭集识别方法,该方法通过对软判决序列在LD模型中理论分布进行期望分析,得出了对不同行重校验向量的理论分布参数,通过对参数的数值分析,给出了实际分布参数与理论分布参数差异性的量化模型,并基于该模型识别出对应校验矩阵。实验结果表明,当码字数量较少时,该方法在多行重校验矩阵中的识别性能优于现有的后验概率识别算法。
In existing LDPC recognition methods over a candidate set,directly using posterior probability to construct models and accumulate results leads to poor recognition performance when the number of codewords is small.Additionally,parameter analysis methods based on distribution fitting using the central limit theorem deviate from the actual distribution under certain conditions,thus failing to achieve higher recognition performance.To address these issues,this paper proposes a distribution parameterbased LDPC recognition method over a candidate set.This method analyzes the theoretical distribution of soft-decision sequences in a likelihood difference model through expectation analysis,deriving theoretical distribution parameters for different row-weight parity-check vectors.By performing numerical analysis on these parameters,a quantitative model for the difference between actual and theoretical distribution parameters is established,and the corresponding parity-check matrix is identified based on this model.Results show that when the number of codewords is small,this method outperforms the existing posterior probability recognition algorithms in terms of recognition performance in multi-row redundant check matrix.
作者
蒲逊
夏明赟
PU Xun;XIA Ming-yun(The 30th Research Institute of China Electronics Technology Group,Chengdu 610041,China)
出处
《中国电子科学研究院学报》
2025年第5期539-545,共7页
Journal of China Academy of Electronics and Information Technology
关键词
LDPC码
闭集识别
软判决
分布参数
low density parity check code
recognition over a candidate set
soft decision
distribution parameter
