New Rateless Sparse-Graph Codes with Dynamic Degree Distribution for Erasure Channels
Xingkai Bao, Jing Li · 2008
A new class of rateless low-density parity-check (LDPC) codes, called the dynamic degree distribution (DDD) codes, are proposed for erasure channels. The parity symbols are generated from a set of degree distributions, one embedded in one another and each optimized for different channel erasure rates. Thus, instead of sticking to a fixed degree distribution for all the channel conditions, the encoder progressively adapts different degree distributions for different channel conditions, such that each distribution is operated in its most-desirable region. Theoretic analysis shows that the new codes perform close to the capacity in a large range of code rates with fairly low complexity. Simulations confirm that these codes can actually outperform the existing raptor rateless codes.