小样本条件下基于矩阵乘法和秩分析的LDPC参数估计方法

LDPC Parameter Estimation by Matrices Product and Rank Analysis Under the Condition of Small Sampling

在非合作通信背景下,利用传统的盲识别算法获取有用信息往往需要大量的截获数据.本文利用少量的截获数据,基于码字空间与其对偶空间的正交性、完整码字比特间的线性相关性和矩阵乘积秩的性质,提出了矩乘秩减算法,在无误码和低误码率情形下恢复了LDPC(Low-Density Parity-Check)长码的码长和起点.仿真实验表明,与传统算法相比,达到同样的识别效果本文算法能够节省至少20%的数据量,且运算量没有明显增加.

In the scenario of non-cooperative communications,usually it takes a large amount of intercepted data for blind identification to obtain useful information with the traditional methods.This paper presents an approach called rank reduction with matrices production to estimate block length and synchronization of long LDPC(Low-Density Parity-Check)codes under the condition of small sampling with noise-free or lower bit-error rate data.Our method is based on the orthogonality of codeword space and its dual space,the linear correlations among the bits in a whole codeword,and the property of rank reduction of matrices.Experimental results show that,our method can save 25%data at least to reach the same identification probability compared with the traditional methods,and the computation has no obvious increasement.