Optimization of Parity-Check Matrices of LDPC Codes

Low-density parity-check (LDPC) codes are widely used in communications due to their excellent practical performance. Error probability of LDPC code under iterative decoding on the binary erasure channel is determined by a class of combinatorial objects, called stopping sets. Stopping sets of small size are the reason for the decoder failures. Stopping redundancy is defined as the minimum number of rows in a parity-check matrix of the code, such that there are no small stopping sets in it. Han, Siegel and Vardy derive upper bounds on the stopping redundancy of general binary linear codes by using probabilistic analysis. For many families of codes, these bounds are the best currently known. In this work, we improve on the results of Han, Siegel and Vardy by modifying their analysis. Our approach is different in that we judiciously select the first and the second rows in the parity-check matrix, and then proceed with the probabilistic analysis. Numerical experiments confirm that the bounds obtained in this thesis are superior to those of Han, Siegel and Vardy for two codes: the extended Golay code and the quadratic residue code of length 48.