Average-entropy variation in iterative decoding of turbo codes and its application

Abstract

A new stopping criterion for turbo codes is proposed. Based on the entropy concept, a metric called average-entropy to measure the average uncertainty of the estimated bits of each iteration is derived. This metric has a close relation to the bit error rate (BER). The average-entropy decreases as BER reduces and vice versa. The proposed stopping criterion stops the iterative algorithm when there is not a, or little, reduction of the average-entropy. Compared with other well-known criteria, this new criterion reduces the average number of iterations while maintaining error correction performance.