AE-BP: Adaptive Erasure Belief Propagation Decoding Algorithm of LDPC Codes

In the decoding process of LDPC codes, variable node messages often fluctuate continuously and thereby become the hindrance to the successful decoding. This paper proposes an adaptive erasure belief propagation (AE-BP) decoding algorithm to reduce the influence of these unreliable messages. The idea behind AE-BP is recording the continuous fluctuant times of variable nodes by introducing a sequence of counters, and then adaptively erasing the messages according to their fluctuant times so that other messages would recover to more precise states. Semi-Gaussian approximation demonstrates adaptive erasure is a good candidate towards offsetting the defect of unreliable messages. Experimental simulations show that, for LDPC codes AE-BP outperforms most decoding algorithms in the literature.