[Paper Review] Aspects of Polar-Coded Modulation

Mathis Seidl, Andreas Schenk+2 more|arXiv (Cornell University)|Nov 29, 2012

Advanced Wireless Communication Techniques6 references13 citations

TL;DR

This paper investigates polar-coded modulation by establishing a theoretical equivalence between polar-coded bit-interleaved coded modulation (BICM) with Gray labeling and multi-level coding (MLC) with set-partitioning (SP) labeling for 4-QAM. It demonstrates that the first stage of a polar code in Gray-labeled BICM is equivalent to SP-labeled MLC, and shows that optimizing frozen channel selection for the specific modulation (e.g., 4-ASK) significantly improves performance over binary-AWGN-optimized codes.

ABSTRACT

We consider the joint design of polar coding and higher-order modulation schemes for ever increased spectral efficiency. The close connection between the polar code construction and the multi-level coding approach is described in detail. Relations between different modulation schemes such as bit-interleaved coded modulation (BICM) and multi-level coding (MLC) in case of polar-coded modulation as well as the influence of the applied labeling rule and the selection of frozen channels are demonstrated.

Motivation & Objective

• To investigate the conceptual and practical equivalence between polar-coded BICM and MLC in higher-order modulation.
• To analyze the impact of labeling schemes (Gray vs. set-partitioning) on polar-coded modulation performance.
• To demonstrate that optimizing frozen channel selection for the target modulation system yields significant performance gains over binary-AWGN-optimized designs.
• To motivate a joint optimization framework for polar codes and modulation that accounts for the specific signal constellation and system operating point.

Proposed method

• Uses the recursive structure of polar codes to relate bit-channel capacities to modulation bit levels.
• Applies the transformation matrix T to convert Gray labeling into set-partitioning (SP) labeling, showing equivalence between BICM and MLC for 4-QAM.
• Derives the equivalence between the first stage of a polar-coded Gray-labeled BICM system and SP-labeled MLC with component codes of half the length.
• Employs numerical simulations to compare bit error rate (BER) performance of polar-coded BICM and MLC under different frozen channel selections.
• Compares performance using rate-distortion bounds and uncoded BPSK as reference points.
• Optimizes frozen channel selection for 4-ASK based on the target Es/N0 operating point, contrasting with binary-AWGN-optimized designs.

Experimental results

Research questions

• RQ1 Is there a theoretical equivalence between polar-coded BICM with Gray labeling and MLC with SP labeling for 4-QAM?
• RQ2 How does the choice of frozen channels affect the performance of polar-coded modulation in higher-order systems?
• RQ3 Can the transformation between Gray and SP labeling be leveraged to unify or optimize polar-coded BICM and MLC designs?
• RQ4 Does optimizing the polar code construction for the actual modulation system (e.g., 4-ASK) outperform binary-AWGN-optimized code design?
• RQ5 What is the impact of interleaving and labeling on the performance of polar-coded modulation?

Key findings

• For 4-QAM, polar-coded BICM with Gray labeling is equivalent to polar-coded MLC with SP labeling when the code length is a power of two and the first stage of the polar code corresponds to the SP-labeled component code.
• The transformation matrix T that maps Gray to SP labeling is identical to the recursive construction matrix of polar codes, establishing a direct structural link between the two schemes.
• Optimizing frozen channel selection for the target 4-ASK system (e.g., at 4 dB Es/N0) yields a BER performance gain of approximately 1.5 dB over a binary-AWGN-optimized code at the same rate.
• The performance of BICM is highly sensitive to frozen channel selection, and the best performance is achieved only when the code is trained for the specific operating point.
• Polar-coded MLC with SP labeling and length-128 component codes achieves performance close to the MLC rate-distortion bound, validating the effectiveness of the joint design.
• The results suggest that joint optimization of polar codes and modulation, including labeling and frozen channel selection, is essential for achieving optimal spectral efficiency in practical systems.