The focus of this chapter is on low-density parity-check (LDPC) codes and iterative decoding methods. Low density parity check codes were originally introduced by Gallager in his doctoral thesis in 1963 [19]. Gallager discussed several iterative decoding algorithms for these codes that in general have very good error-correcting capabilities. Later, Tanner [32] in 1981 gave a graphical interpretation to these codes as a bi-partite graph whose incidence matrix is the parity check matrix of the LDPC code, and showed how to view the iterative decoding algorithms as message-passing algorithms on these graphs.
A construction technique is proposed for low-density parity-check (LDPC) codes based on finite Euclidean geometries EG(m,2s). These codes are shown to be regular Gallager codes with Tanner graphs of girth eight. The minimum distance of these codes is shown to be lower-bounded by 2m. The codes are also amenable to an efficient partly parallel decoder implementation, which may be used in conjunction with the turbo decoding message passing (TDMP) algorithm for LDPC decoding. Finally, simulation results show that these codes have very good error-correcting performance
LDPC codes have excellent performance when iteratively decoded based on sparse parity-check matrices over both the AWGN channel and the erasure channel. In this paper, we propose a simple burst-correction decoding scheme for cyclic LDPC codes based on the same sparse circulant parity-check matrices used to perform iterative decoding. Although the proposed scheme may not achieve the full burst-correcting capability of the codes, it is rather simple and fast. We also show that the full burst-correcting capabilities of some codes constructed from Euclidean and projective geometries are very close to the Reiger upper bound