Abstract:

Designing a realizable maximum likelihood (ML) decoder for low-density parity-check (LDPC) codes is always a challenging work over an additive white Gaussian noise (AWGN) channel. Although Maxwell decoder is well known for its excellent performance over the binary erasure channel (BEC), it seems that generalizing this algorithm for other channels is difficult. This paper introduces an idea called channel transformation which could realize the conversation between two different channels. A Maxwell decoder is applied to an AWGN channel. In terms of this method, and a joint decoder-BP-Maxwell (BM) decoder is proposed which combines a belief propagation (BP) decoder and a Maxwell decoder, to reduce the gap to the ML decoder in performance. Simulation results show that the BM decoding algorithm could break most small trapping sets to accomplish a lower frame error rate (FER). Moreover it also could eliminate most of the smallscale errors compared with a BP decoder.