Modeling the bit-level stochastic correlation for turbo decoding

After passing a systematic bit through a turbo encoder, the encoding process will introduce some extent of correlation between a systematic bit and its associated parity bits. However, this correlation is neglected in the subsequent turbo decoding process so as to reduce its computational complexity. In this paper, we try to explore the feasibility of modeling the bit-level stochastic correlation for the iterative turbo decoding. By properly adjusting the parameter of the correlation model, we can approximate various degrees of the underlying correlation within the received codewords. Reduction in bit error rate (BER) then may benefit from a more accurate capture of the correlation information at the cost of requiring only a small additional computation complexity. Experimental results indicate that incorporating the correlation model into the turbo decoding process can achieve better BER performance than that of conventional turbo decoders over AWGN channels.