The New Multi-Edge Metric-Constrained PEG/QC-PEG Algorithms for Designing the Binary LDPC Codes With Better Cycle-Structures.
Xuan He,Liang Zhou,Junyi Du
DOI: https://doi.org/10.1109/tcomm.2017.2753805
2016-01-01
Abstract:To obtain a better cycle-structure is still a challenge for the low-density parity-check (LDPC) code design. This paper formulates two metrics firstly so that the progressive edge-growth (PEG) algorithm and the approximate cycle extrinsic message degree (ACE) constrained PEG algorithm are unified into one integrated algorithm, called the metric-constrained PEG algorithm (M-PEGA). Then, as an improvement for the M-PEGA, the multi-edge metric-constrained PEG algorithm (MM-PEGA) is proposed based on two new concepts, the multi-edge local girth and the edge-trials. The MM-PEGA with the edge-trials, say a positive integer $r$, is called the $r$-edge M-PEGA, which constructs each edge of the non-quasi-cyclic (non-QC) LDPC code graph through selecting a check node whose $r$-edge local girth is optimal. In addition, to design the QC-LDPC codes with any predefined valid design parameters, as well as to detect and even to avoid generating the undetectable cycles in the QC-LDPC codes designed by the QC-PEG algorithm, the multi-edge metric constrained QC-PEG algorithm (MM-QC-PEGA) is proposed lastly. It is verified by the simulation results that increasing the edge-trials of the MM-PEGA/MM-QC-PEGA is expected to have a positive effect on the cycle-structures and the error performances of the LDPC codes designed by the MM-PEGA/MM-QC-PEGA.