Joint Decoding of LDPC Codes and Finite-State Channels via Linear-Programming

TL;DR

This paper extends linear-programming decoding to joint decoding of LDPC codes and finite-state channels, providing theoretical bounds, an efficient iterative solver, and demonstrating promising performance and complexity advantages.

Contribution

It introduces a novel LP joint-decoding framework for FSCs, defines JD-PCWs, and develops an iterative solver with performance comparable to LP decoding and complexity similar to turbo equalization.

Findings

Provides a rigorous definition of JD-PCWs for error analysis

Develops an efficient iterative solver with convergence guarantees

Achieves performance close to joint LP decoding with TE-like complexity

Abstract

This paper considers the joint-decoding (JD) problem for finite-state channels (FSCs) and low-density parity-check (LDPC) codes. In the first part, the linear-programming (LP) decoder for binary linear codes is extended to JD of binary-input FSCs. In particular, we provide a rigorous definition of LP joint-decoding pseudo-codewords (JD-PCWs) that enables evaluation of the pairwise error probability between codewords and JD-PCWs in AWGN. This leads naturally to a provable upper bound on decoder failure probability. If the channel is a finite-state intersymbol interference channel, then the joint LP decoder also has the maximum-likelihood (ML) certificate property and all integer-valued solutions are codewords. In this case, the performance loss relative to ML decoding can be explained completely by fractional-valued JD-PCWs. After deriving these results, we discovered some elements were equivalent to earlier work by Flanagan on LP receivers. In the second part, we develop an efficient iterative solver for the joint LP decoder discussed in the first part. In particular, we extend the approach of iterative approximate LP decoding, proposed by Vontobel and Koetter and analyzed by Burshtein, to this problem. By taking advantage of the dual-domain structure of the JD-LP, we obtain a convergent iterative algorithm for joint LP decoding whose structure is similar to BCJR-based turbo equalization (TE). The result is a joint iterative decoder whose per-iteration complexity is similar to that of TE but whose performance is similar to that of joint LP decoding. The main advantage of this decoder is that it appears to provide the predictability of joint LP decoding and superior performance with the computational complexity of TE. One expected application is coding for magnetic storage where the required block-error rate is extremely low and system performance is difficult to verify by simulation.