• DocumentCode
    802900
  • Title

    Algebraic construction of sparse matrices with large girth

  • Author

    O´Sullivan, Michael E.

  • Author_Institution
    Dept. of Math. & Stat., San Diego State Univ., CA
  • Volume
    52
  • Issue
    2
  • fYear
    2006
  • Firstpage
    718
  • Lastpage
    727
  • Abstract
    In this correspondence, we present a method for constructing sparse matrices that have a compact description and whose associated bipartite graphs have large girth. Based on an arbitrary seed matrix of nonnegative integers a new matrix is constructed which replaces each entry of the seed matrix with a sum of permutation matrices. Algebraic conditions that lead to short cycles in the associated bipartite graph are analyzed and methods to achieve large girth in two special cases are presented. In one, all the permutation matrices are circulants; in the other they are all affine permutation matrices. When used to define a low-density parity-check (LDPC) code the compact description should lead to efficient implementation and the large girth to good error correction performance. The method is adaptable to a variety of rates, and a variety of row and column degrees
  • Keywords
    error correction codes; graph theory; parity check codes; sparse matrices; LDPC; algebraic construction; arbitrary seed matrix; associated bipartite graph; circulant permutation matrix; error correction performance; large girth; low-density parity-check code; nonnegative integer; sparse matrix; Belief propagation; Bipartite graph; Decoding; Error correction codes; Floors; Information theory; Parity check codes; Sparse matrices; Sum product algorithm; Upper bound; Bipartite graph; low-density parity-check (LDPC) codes;
  • fLanguage
    English
  • Journal_Title
    Information Theory, IEEE Transactions on
  • Publisher
    ieee
  • ISSN
    0018-9448
  • Type

    jour

  • DOI
    10.1109/TIT.2005.862120
  • Filename
    1580807