Enhanced Cross-Entropy Based Stopping Criteria at Low Signal-to-Noise Ratio Regions

Abstract

This paper presents an enhancement of Cross-Entropy (CE)-based stopping criteria at a low signal-to-noise ratios (SNRs) region. The existing CE-based stopping criteria fail to stop the iteration and cause the maximum iteration at low SNRs. Intuitively, this work enhances the CE stopping criterion by combining with the Average-Entropy (AE) stopping criteria. The work also proposes a set of a stopping rule to distinguish the SNR between low and high region. These approaches are then thoroughly analyses in terms of the required average iteration number (AIN), as well as its bit-error-rate (BER) performance. The simulation results show that the enhanced CE able to reduce the AIN compared to existing CE-based stopping criteria while the BER is maintained. Hence, the enhanced CE-based stopping criterion make the decoder more efficient, in terms of reduction in the time delay while increasing turbo codes performance.