Performance of iterative turbo decoding with the hypothesis of generalized Gaussian distribution for extrinsic values in AWGN and fading channels

A new strategy for iterative turbo decoding is proposed, where a class of symmetric Gaussian-like probability distributions are used to estimate the conditional probabilities of extrinsic values generated from branch detectors. A specific probability with a unique parameter is finally determined based on a proposed likelihood criterion of minimum estimated error bits for any detector in each iterative decoding. In this sense, the optimum distributions of the extrinsic values of iterative decoding are unveiled over AWGN, Rician and Rayleigh fading channels. The simulation results show that the decoding with this new technique has superior error performance to the traditional approach based on the conventional Gaussian law for the extrinsic values under the same channel conditions