A New Stopping Criterion for Turbo Decoders Based on the Minimization of Error Probability

This paper proposes a new criterion to stop the iterations in turbo decoders, based on the minimization of the probability of erroneous decisions, hence the name MinPe. The criterion analyses the evolution of Log-Likelihood Ratio (LLR) and compares the distance between the distributions of the LLR's negative and positive values with a threshold depending on the average of positive LLR's values. The iterative decoding process is stopped if the distance is greater than the threshold. The performances are analyzed in terms of Average Number of Iterations (ANT) and Bit Error Rate (BER) and compared with those of other classical criterions.