Low-Density Parity-Check (LDPC) codes are a class of linear blockcodes. Due to their powerful error-correcting ability and Shannoncapacity approaching performance when codeword length is longenough, LDPC codes have become a hot topic in wirelesscommunication channel coding area and widely adopted in manywireless communication protocols.Recently, their counterpart nonbinary LDPC codes constructed overGalois field of size q have shown their potential in improving the codinggain especially at moderate code lengths. Furthermore, nonbinary LDPCcodes can provide high data transmission rate because they areappropriate for combining with high order modulation.The main obstacle for application of nonbinary LDPC codes is thehuge computational complexity. Traditional belief propagation baseddecoding algorithms can provide good decoding performance at the costof high decoding complexity. Many efforts have already been done toreduce the computational complexity. Among these algorithms, majoritylogic based message-passing algorithm stands out for their significantreduction in computational complexity. In this research work, we focuson improving this majority logic based decoding algorithm and providingefficient trade-off between complexity and performance. By updatingreliability measure of prediction from check node to variable node, wecan provide more precise information for decoding.Flooding scheduling scheme is a general scheduling method forbelief propagation and majority logic based algorithm. Nevertheless, recent studies have shown that dynamic scheduling using the latestavailable information is superior to the flooding scheduling. Whileexisting works on informed dynamic scheduling focus on beliefpropagation based algorithm, in this research we devise a novel methodwhich is appropriate for majority logic decoding of non-binary LDPCcodes. By utilizing the stability information of variable and check nodes,we propose a new ordering metric to select the messages for propagationin each iteration. Simulation results verify that our approach can achievebetter performance compared with previous works and speed up theconvergence. |