On the Robustness of Average-Entropy Stopping Criterion in Turbo Iterative Decoding

Abstract

Average-entropy (AE) is an efficient early stopping criterion that is able to terminate early in a varying signal-tonoise ratio (SNR) environment while suffering very little performance degradation. However, the performance of AE is only based on two code structures and depends on correct SNR; hence, the robustness of AE is still unknown, especially for other code structures and in incorrect SNR estimation. Thus, this paper presents the robustness test and analysis of AE based on the following parameters: frame size, code structure, channel reliability, and code rate. From the analysis, AE is capable to achieve good average iteration number performance for SNR ≤ 0.5 dB while maintaining the BER performance in correct SNR estimation. AE can also operate well in constant estimated SNR= 0 dB. However, AE suffers BER degradation for SNR > 0.5 dB both in correct SNR estimation and constant estimated SNR= 1 dB.