Exploiting redundancies to improve performance of LT decoding

Rateless codes are a class of codes without a predefined number of encoding symbols. Fountain codes are the first such codes. Luby Transform codes, designed by Michael Luby, are examples of these codes. The accepted and efficient decoding algorithm for these dense bipartite codes is the belief propagation (BP) algorithm as opposed to Gaussian elimination algorithm. LT codes are more efficient as the number of symbols grow. However for streaming purposes, the number of source symbols needs to be constrained while at the same time the probability of success be maintained. In this paper we propose a BP decoding enhancement algorithm using redundancies in the already received encoding symbols to improve the probability of performance of LT decoding. We show that in the limit our proposal achieves up to 15% performance improvement in decoding throughput with similar channel conditions. We emulate our proposal in software and the results follow that predicted by theoretical analysis