低错误平层数列分割移位低密度奇偶校验码构造算法

发布时间：2019-05-31 16:55

【摘要】：为降低LDPC(低密度奇偶校验码)码错误平层,提出一种基于环分类搜索的APPS-LDPC(数列分割移位的LDPC)码构造算法。该算法具有码长、码率和列重的任意可设性,同时该类码的Tanner图围长至少为8。循环移位因子可以通过简单的代数表达式描述,从而降低内存需求。仿真结果表明,当误码率达到10-5时,APPS-LDPC码(496,248)相对于PEG-LDPC(渐进边增长LDPC)码获得了约1.9 d B的性能提升;随着信噪比的升高,两条译码性能曲线之间的差距将更大。此外,列重为3的APPS-LDPC码(6144,5376)在信噪比4.6 d B以后并未出现明显的错误平层。该构造算法与PS-LDPC码相比,在误码率达到10-8时大约获得0.25 d B增益;与围长为4和6的PEG构造算法相比,在错误平层区域其译码性能极优;同时相较于此两者,其构造复杂度和耗时也展现出一定优势。通过基于Tanner图的诱捕集分析方法,统计APPS-LDPC码(496,248)中由8环组成的部分小型诱捕集并不存在,从而证明了其错误平层降低的原因。
[Abstract]:In order to reduce the error flat layer of LDPC (low density parity check code) code, a APPS-LDPC (sequence division shift LDPC) code construction algorithm based on ring classification search is proposed. The algorithm has arbitrary setability of code length, code rate and column weight, and the Tanner graph circumference of this kind of code is at least 8. Cyclic shift factor can be described by simple algebra expression, thus reducing memory requirement. The simulation results show that when the bit error rate reaches 10-5, the performance of APPS-LDPC code (496248) is about 1.9dB higher than that of PEG-LDPC (progressive edge growth LDPC) code. With the increase of signal-to-noise ratio (SNR), the gap between the two decoding performance curves will be even greater. In addition, there is no obvious error flat layer in the APPS-LDPC code (6144, 5376) with column weight 3 after the SNR of 4.6dB. Compared with the PS-LDPC code, the proposed algorithm achieves about 0.25dB gain at the BER of 10-8, and the decoding performance of the proposed algorithm is excellent compared with the PEG construction algorithm with girth 4 and 6, and the decoding performance of the proposed algorithm is very good compared with the PEG construction algorithm with girth 4 and 6. At the same time, compared with the two, its structural complexity and time-consuming also show some advantages. By using the entrapment analysis method based on Tanner graph, some small entrapment sets composed of eight rings in APPS-LDPC code (496248) do not exist, and the reason for the decrease of error flat layer is proved.
【作者单位】：国防科技大学电子科学与工程学院;中国电子科技集团公司第二十九研究所;
【基金】：国家自然科学基金资助项目(61372098,61501479)
【分类号】：TN911.22

【相似文献】