Importance sampling for loop-free decoding trees with application to low-density parity-check codes

摘要

An importance sampling scheme for the decoding of loop-free decoding trees is proposed. It is shown that this scheme is asymptotically efficient and that, for an arbitrary tree and a given estimation precision, the number of simulation runs needed is inversely proportional to the standard deviation of the channel noise. This work has its application to the decoding of low-density parity-check codes.