Binary and non-Binary Turbo Codes for the Compression of Correlated Sources Transmitted through Error-Prone Channels

Abstract

In this paper, we study the performances of several non-binary turbo-codes transmitted over Gaussian and binary symmetrical channels. We prove that duo-binary codes constitute the best trade-off between error-correction capacity and decoding complexity. We also present an efficient method for the calculation of the transition probability in the Max-Log-MAP decoding algorithm, for the case where compressed binary sequences are transmitted in the presence of noise. We particularly address the problem of distributed source coding where the side information and the parity sequence experience different channels. We also calculate the theoretical compression bound for a noisy transmission and estimate the average compression rate required for acceptable decoding performances.