Near-Log-MAP Decoding for Turbo Product Codes

Turbo product codes (TPCs) have been studied since 1994. However, existing decoding algorithms for TPCs are generally based on the algorithm presented by Pyndiah [1], which adopts the idea of MAX-Log-MAP algorithm and neglects the channel side information (CSI). In this paper, we present and investigate the "M-Log-MAP" decoding algorithm that considers multiple codewords in calculating the extrinsic information. It is shown that the algorithm can improve the decoding performance to the near-Log-MAP decoding with a small M (≈ 4). The complexity of the decoding algorithm is also analyzed.