LDPC码最小和译码算法的整数量化

低密度奇偶校验码(low density parity check codes, LDPC)以其接近香农极限的性能和相对简单的译码结构得到信道编码界的广泛关注。对LDPC码的最小和算法进行了深入地研究,通过多种方法量化译码时的初始消息,最终使得每次迭代的校验消息与变量消息都变为整数,实现了基于整数运算的最小和译码算法,并进行了对比分析。仿真表明,量化后的最小和算法中的所有变量都用固定长度的整数表示,因而便于硬件实现,在其译码性能比和积译码(sum product decoding, SP)性能下降不大的情况下大大提高了译码速度;平均互信息越大的量化方法,其量化分层电平也越佳;最大平均互信息量化下的最小和译码算法性能最好,最大平均互信息量化是一类能最大可能获得信源信息条件下的最佳量化方法,且不增加译码复杂度。

Min sum decoding algorithm of LDPC codes based on integer quantisation

Low-density parity-check (LDPC) codes become popular for its near Shannon limit performance and relatively simple decoding structure. The min sum decoding algorithm of LDPC codes is studied detailedly. And the variable message and check message of each iteration become integers through a variety of methods quantifying the initial message. Then the integer arithmetic based on the min sum decoding algorithm is realized. Finally, these algorithms are compared and analyzed. Simulation results show that all variables of the min sum decoding algorithm are fixed length integers after quantification. It is easy for hardware implementation. And the decoding time is greatly shortened under the condition that the decoding performance degrades less than the sum product (SP) decoding. The greater the average mutual information of the quantitative method is, the better the quantized level will be. The min sum decoding algorithm of the maximum average mutual information quantization has the best performance. The maximum average mutual information quantization is the best quantitative method to keep the source information as much as possible, and it does not increase the complexity of decoding.