中文摘要：

介绍了低密度奇偶校验码（LDPC）的几种常用译码算法及其优缺点,特别用密度进化理论分析了归一化置信传播（Normalized BP-based）和偏移置信传播算法（Offset BP-based）的外信息概率分布和演化。基于此,分别针对Normalized BP-based和Offset BP-based算法提出了广义互信息理论（Generalized Mutual Information）及其计算公式,同时提出了改进的因子自适应LDPC译码算法,在每一次译码过程中通过一维搜索,可以获得一个最佳的修正因子,该因子能够最大化广义互信息,从而获得最佳的译码性能。分析和仿真数据表明,提出的因子自适应算法比传统的算法具有更好的性能。

英文摘要：

This paper first presents several commonly-used decoding algorithms for the well-know low-density-parity-check（LDPC） code,then analyzes its pros and cons. The density evolution theory is employed to trace the probability density function of the extrinsic information in the iterative decoding for the Normalized BP-based and Offset BP-based algorithms. This paper further describes the concept of generalized mutual information and proposes two formulas for N-BP and O-BP-based algorithms respectively. And upon this,a coefficient-adaptive decoding algorithm is proposed. In each decoding iteration,the corresponding optimal correction coefficient can be obtained via one dimensional global search,this could also maximize the generalized mutual information,i.e.,guarantee the best decoding performance. Analysis and simulation results indicate that the proposed coefficient-adaptive algorithm is better than the traditional algorithms.