Adaptive decoding algorithm based on multiplicity order of candidate sequences for TPC

Abstract

It is known that Turbo Product Codes (TPC) can be nearly optimally decoded by Chase-II algorithm, in which the Least Reliable Bits (LRBs) are chosen empirically to keep the size of the test patterns (sequences) relatively small and to reduce the decoding complexity. While there are also other adaptive techniques, where the decoder's LRBs adapt to the external parameter of the decoder like signal SNR level, a novel adaptive algorithm for TPC based on the statistics of an internal variable of the decoder itself is proposed in this paper. Different from the previous reported results, it collects the statistics of multiplicity order of the candidate sequences, i.e., the number of the same candidate sequences with the same minimum squared Euclidean distance returned from the decoding of test sequences. It is shown by Monte Carlo simulations that the proposed adaptive algorithm has only about 0.03dB coding loss but the average complexity of the proposed algorithm is about 45% less compared with Pyndiah's iterative decoding algorithm using the fixed LRBs parameter.