Low complexity early stopping structure for belief propagation decoder

Abstract

Belief propagation (BP) decoder is a well known iterative decoder used for decoding many error correction code families. BP decoder without early stopping structure uses a fixed iteration number to end iterative decoding process. But decoder may be converged before iteration number reaches this fixed limit.~In these case, decoder performs a redundant process. Therefore, stopping the iterations is essential to keep computational burden as low as possible when decoding is successful. With this perspective, a low complexity early stopping structure for belief propagation decoders is proposed with this thesis.~In contrast to previous early stopping methods in literature, proposed early stopping structure only uses small amount of log-likelihood ratio (LLR) messages and tracks only sign alterations of them. Proposed structure is applied to both polar and Luby transform (LT) codes and can be easily applied to error correction codes use BP as decoder.~Performance parameters are compared with simulation works and hardware description language (VHDL) implementations.~Results illustrate that proposed approach significantly reduces the computational complexity and required hardware resources, also throughput values are increased compared to previous counterparts in literature.