Adaptive early-stopping threshold for LTE turbo decoder

Turbo coding has been adopted by 3GPP-LTE standard for error correction. However, decoding with a fixed number of iterations may lead to excessive processing, penalizing battery-operated terminals while fewer iterations are sometimes sufficient. Early-stopping criteria can stop the iterative process when a certain confidence threshold has been achieved. However, practical implementations use a fixed threshold. In this paper we propose an adaptive approach based on the block size and coding rate to select the best threshold for two state-of-the-art early-stopping techniques. We show gains up to 38% in average number of iterations with respect to a fixed-threshold approach and higher gains with respect to a fixed-iterations approach without degradation under the LTE performance constraints. Moreover, our approach does not require SNR knowledge.