Rate-adaptive constellation shaping for near-capacity achieving turbo coded BICM

In this paper the problem of constellation shaping is considered. Mapping functions are designed for a many-to-one signal shaping strategy, combined with a turbo coded Bit-interleaved Coded Modulation (BICM), based on symmetric Huffman codes with binary reflected Gray-like properties. An algorithm is derived for finding the Huffman code with such properties for a variety of alphabet sizes, and near-capacity performance is achieved for a wide SNR region by dynamically choosing the optimal code rate, constellation size and mapping function based on the operating SNR point and assuming perfect channel quality estimation. Gains of more than 1dB are observed for high SNR compared to conventional turbo coded BICM, and it is shown that the mapping functions designed here significantly outperform current state of the art Turbo-Trellis Coded Modulation and other existing constellation shaping methods.