A Systematic Constellation Design for BICM Systems With Geometric Shaping

We investigate the application of geometric constellation shaping for standard bit-interleaved coded modulation (BICM) systems employing Gray labeling and high-order quadrature amplitude modulation (QAM). We focus on the non-uniform Gaussian-like distributions, and analyze their performance in terms of the mutual information as well as more practical bit-wise mutual information (BMI). We also propose a new systematic constellation design approach based on truncated Gaussian distribution, where a single parameter balances the two extreme constellations represented by uniform and Gaussian distributions. We also evaluate the peak-to-average power ratio (PAPR) of the resulting constellations, where a good trade-off between the BMI and PAPR can be achieved by optimizing the associated parameter. Simulation results based on the conventional Gray-labeled BICM with binary turbo codes show that a noticeable shaping gain can be achieved by the proposed constellation design.