Analysis of iterative decoding of turbo codes

Abstract

An analytical model for the extrinsic information of iterative max-log-MAP decoding of turbo codes is established based on a maximum probability approximation. The extrinsic information can be expressed as a Gaussian random variable with a mean and variance that are linear combinations of the received symbols. An important measure, called the effective weight, which is related to the pairwise error probability, is defined to explain the iterative decoding effect on a given code sequence. The evaluation of the pairwise error probability performance through decoder iterations can be calculated based on this model. The proposed approach can serve as a tool for the analysis of iterative max-log-MAP decoding.