Utilization of 2-D Markov source correlation using block turbo codes

Abstract

This paper proposes a joint source-channel coding technique using high rate block turbo codes where two-dimensional (2-D) source correlation of binary Markov source is well exploited. A modified Bahl-Cocke-Jelinek-Raviv (BCJR) algorithm is used in both horizontal and vertical decoders to exploit the 2-D source correlation. Based on extrinsic information transfer (EXIT) chart analysis and trajectory evaluation, we found out that the log-likelihood ratio (LLR) correlation for the block turbo codes increases as the source correlation becomes stronger and this prevents iterative decoding from convergence. A memory-1 accumulator with random interleaver is added to the design to minimise the LLR correlation and to achieve better performance when the source correlation is very strong. Further improvement can be achieved by replacing the memory-1 accumulator with a longer memory code to further reduce the statistical dependency of the LLR. It is shown that the proposed technique can achieve significant improvement over the standard system where the source correlation is not exploited