Joint source-channel decoding of IRA code for hidden Markov source
Saikat Majumder, Shrish Verma · 2012
We present the design of a joint source-channel decoder by modifying the IRA code decoder in order to exploit the correlation characteristics of hidden Markov sources. The basic idea is to add factor graph for forward/backward algorithm, which models the Markov source, to the bipartite graph of IRA decoder. We assume that receiver has no apriori knowledge of the correlation characteristics of the source. The joint decoder uses an iterative algorithm which passes message to and fro between basic IRA decoder and hidden Markov nodes to estimate the transmitted message. The proposed scheme leads to significantly improved performance compared to system in which source correlation statistics are not utilized and avoids the need to perform a separate data compression prior to channel coding and transmission.