改进的低密度校验码的定点译码实现

Improved decoding approach of low-density parity-check codes with finite precision

为了解决低密度校验码(LDPC)的并行译码实现方法在定点设计中出现的迭代发散问题,提出了一种改进的并行译码实现方法,同时分析了改进后加速迭代收敛的原因,并且给出在硬件实现中的流水线结构.通过消除定点设计时,由有限精度效应导致的迭代过程中运算值不满足函数输入条件、无法正确处理的异常状态,使得迭代迅速收敛,得出正确的译码结果;且硬件实现资源与原有的并行译码实现方法相比,没有增加过多的控制逻辑和运算复杂度.仿真结果表明,采用该实现方法的LDPC定点译码设计在AWGN信道下得到了良好的性能,特别在高码率下,与前-后向实现方法相比性能差异可以忽略,且硬件复杂度更低.

An improved parallel decoding approach of sum-product algorithm for low-density parity-check codes （LDPC） was presented in order to solve the problem of divergence in decoding iteration with finite precision,. In addition, the reason for improving the convergence speed was discussed detailedly and the pipeline structure was provided. By eliminating the unexpected state in which operation data didn＇t satisfy input rules due to the finite precision effect thus could not be preceded correctly, the iteration convergence was improved efficiently and led to decoding with no error. Compared with the original parallel approach, this improvement don＇t involve too much control logic and computational complexity. Simulation shows that the proposed decoding approach over additive white Gaussian noise（AWGN） channel achieves good performance, especially at high code rate, it suffers negligible performance loss compared with the forward-backward approach while reducing the operation complexity effectively.