Weight distributions and constructions of low-density parity-check codes.

摘要

Low-density parity-check (LDPC) codes are currently the most promising coding technique to achieve the Shannon capacities for a wide range of channels. These codes were first discovered by Gallager in 1962 [15] and then rediscovered in late 1990's [28, 37]. Ever since their rediscovery, a great deal of research effort has been expended in design, construction, encoding, decoding, performance analysis, generalizations, and applications of LDPC codes.;This research is set up to investigate two major aspects of LDPC codes: weight distributions and code constructions. The research focus of the first part is to analyze the asymptotic weight distributions of various ensembles. Analysis shows that for generalized LDPC (G-LDPC) and doubly generalized LDPC (DG-LDPC) code ensembles with some conditions, the average minimum distance grows linearly with the code length. This implies that both ensembles contain good codes. The effect of changing the component codes of the ensemble on the minimum distance is clarified. The computation of asymptotic weight and stopping set enumerators is improved.;Furthermore, the average weight distribution of a multi-edge type code ensemble is investigated to obtain its upper and lower bounds. Based on them, the growth rate of the number of codewords is defined. For the growth rate of codewords with small linear, logarithmic, and constant weights, the approximations are given with two critical coefficients. It is shown that for infinite code length, the properties of the weight distribution are determined by its asymptotic growth rate.;The second part of the research emphasizes specific designs and constructions of LDPC codes that not only perform well but can also be efficiently encoded. One such construction is the serial concatenation of an LDPC outer code and an accumulator with an interleaver. Such construction gives a code called an LDPCA code. The study shows that well designed LDPCA codes perform just as well as the regular LDPC codes. It also shows that the asymptotic minimum distance of regular LDPCA codes grows linearly with the code length.