Abstract
Incremental decoding (ID) [1] with informed dynamic scheduling (IDS) [2] (IDIDS [3]) can well balance the error-rate performance, decoding complexity and the overhead, and hence is an attractive solution for rateless codes transmitted over binary symmetric channel (BSC). However, when AWGN (additive white Gaussian noise) channel is considered, the incrementally received symbols may not be immediately utilized, and the error-rate performance cannot be maintained. Moreover, using the stopping criterion designed for BSC channel, IDIDS may terminate the decoding process too early, and cannot efficiently use the channel information contained in the received codeword. In this paper, for each decoding attempt, the proposed decoder follows the conventional flooding schedule in the first iteration such that the incrementally received symbols can immediately propagate their channel values. After the first iteration, the schedule of the proposed decoder is switched to IDS. When the decoding convergence is slowing down, the schedule can be switched back to flooding for one iteration, and the messages in the code graph can be refreshed. Then the decoding schedule return to IDS again. The switch can be done for several times before the decoding process is accomplished. Using this decoding strategy, we can achieve improvement in BER performance for both Luby transform (LT) codes and Raptor codes.