Abstract

This paper presents a new method for stopping the iterative turbo decoding. First, a bit-level convergence test using the cross-entropy analyses is used to select non converged bits and establish a simple and effective stopping rule. Next, an adaptive approach is used to compute a scaling factor for normalizing the extrinsic information of the previously selected bits. The extra coding gain obtained from this normalization can compensate for the performance degradation of the stopping rule. The simulation results of the proposed stopping criterion show an interesting application in a hybrid automatic repeat request systems with turbo coding scheme, where the decoding complexity can be fairly reduced.
Simulation results of the proposed criterion, in comparison with previously published stopping rules, were presented for illustrating the adaptive termination according to a changing SNR environment.