There exist two criterions for the existing LDPC decoding algorithms at node-processing
majority-logic processing and fully processing.The algorithms can't make flexible tradeoffs between performance and complexity
since their threshold parameters are usually set to be unvaried.This paper first presents a generalized threshold-function
which can apply to most of the existing binary LDPC decoding algorithms to control the participating-nodes in the iterations by adjusting the threshold parameters.Then a new non-uniform quantization decoding algorithm combined with syndrome and constellation mapping information is presented
which can further reduce the complexity and memory load.Simulation results show that the presented algorithm can obtain excellent decoding performances with very low quantization level over the Rayleigh fading channel.Furthermore
it is shown that only 30% variable nodes are involved in the iterations when combined with the presented threshold-function