Belief propagation with gradual edge removal for Raptor codes over AWGN channel

A novel belief propagation algorithm with gradual edge removal (ERBP) is proposed for Raptor codes over additional white gaussian noise (AWGN) channel. Specifically, the variable nodes with sufficiently high confidence and the corresponding edges are removed gradually from the Tanner graph during the BP iterations so as to reduce the decoding complexity. Then considering the intrinsic nature of incremental redundancy of Raptor codes, we extend the ERBP algorithm into a progressive manner, i.e., the so-called PERBP algorithm, which not only conducts edge removal operation but also makes use of the decoding states from the previous decoding attempts. Simulation results show that compared with the conventional BP algorithm, both ERBP and PERBP algorithms provide drastically reduced decoding complexity without much loss of coding performance.