A Systematic and Low-Complexity Constellation Shaping Design for Rate-Adaptive BICM Systems Based on Nonuniform QAM

In this work, we propose a systematic shaping design based on nonuniform quadrature amplitude modulation (NU-QAM) constellations for rate-adaptive bit-interleaved coded modulation (BICM) systems. Since our main interest lies in low-complexity signal design and high bandwidth efficiency, we focus on one-dimensional nonuniform constellations (1-D NUCs). In the proposed approach, the constellation points are first generated according to a truncated Gaussian distribution specified by a single parameter, and then a novel Gray labeling strategy with many-to-one mapping is applied. This scheme enables a systematic and flexible constellation design in terms of the number of constellation points and its geometrical distribution. Simulation results using turbo codes show that the proposed constellation achieves noticeable shaping gains over the conventional uniformly-spaced QAM across a wide range of SNRs and is almost comparable to the 1-D NUCs adopted by the ATSC 3.0 standard. The major practical advantage of the proposed shaping scheme is its lower demapping complexity due to the reduced number of constellation points. Our numerical analysis indicates that, compared to the equivalent NUC of ATSC 3.0, the proposed constellation with 1024 bit labels successfully reduces the complexity associated with the demapping process by 69% and 41% for code rates of 2/5 and 3/5, respectively, at the cost of a slight loss in shaping gain.