[Paper Review] On the Joint Error-and-Erasure Decoding for Irreducible Polynomial Remainder Codes

Jiun-Hung Yu|arXiv (Cornell University)|Feb 24, 2012

Coding theory and cryptography8 references3 citations

TL;DR

This paper proposes two novel fixed-transform joint error-and-erasure decoding algorithms for irreducible polynomial remainder codes, leveraging GCD-based computations to efficiently decode both errors and erasures. The second approach, which is new even for Reed-Solomon codes, offers lower computational complexity and full compatibility with existing error-only decoding methods.

ABSTRACT

A general class of polynomial remainder codes is considered. Such codes are very flexible in rate and length and include Reed-Solomon codes as a special case. As an extension of previous work, two joint error-and-erasure decoding approaches are proposed. In particular, both the decoding approaches by means of a fixed transform are treated in a way compatible with the error-only decoding. In the end, a collection of gcd-based decoding algorithm is obtained, some of which appear to be new even when specialized to Reed-Solomon codes.

Motivation & Objective

• To extend existing error-only decoding algorithms for irreducible polynomial remainder codes to handle both errors and erasures simultaneously.
• To develop decoding methods that are fully compatible with existing error-only decoding implementations for practical deployment.
• To reduce decoding complexity by avoiding repeated basis re-computation during interpolation, especially in the presence of erasures.
• To provide new, efficient GCD-based decoding algorithms that work for variable-degree moduli, enabling flexible code length and rate design.
• To present a novel decoding approach for Reed-Solomon codes that improves on prior joint decoding methods in terms of computational efficiency.

Proposed method

• Proposes Fixed-Transform Approach I, which uses a transformed version of the received polynomial $\hat{Y}(x)$ and applies a Partial GCD Algorithm I to compute $s(x), t(x), r(x)$ satisfying $r(x) = s(x)\tilde{M}_n(x) + t(x)E(x)$.
• Employs Fixed-Transform Approach II, which uses $\tilde{M}_n(x)$ and $Y(x)$ as inputs to a Partial GCD Algorithm II, reducing input polynomial degrees and thus lowering computational cost.
• Utilizes the Extended GCD Algorithm with loop invariants ensuring $r(x) = s(x)\tilde{M}_n(x) + t(x)E(x)$ throughout iterations, with termination conditions based on degree thresholds.
• Reconstructs the original message polynomial $a(x)$ via two recovery methods: $a(x) = r(x)/(t(x)\Lambda_\rho(x))$ or $a(x) = t(x)\hat{Y}(x) \bmod M_n(x) / (t(x)\Lambda_\rho(x))$ in Approach I, and $a(x) = r(x)/t(x)$ or $a(x) = t(x)Y(x) \bmod \tilde{M}_n(x) / t(x)$ in Approach II.
• Applies degree-weighted distance analysis to ensure minimum distance $d_{\text{minD}} > N - K$, guaranteeing error-correcting capability under the given assumptions.
• Introduces a condition (34) that ensures the correctness of the GCD-based decoding process, with decoding failure declared if $\deg a(x) \geq K$ or if polynomial division fails.

Experimental results

Research questions

• RQ1 Can efficient joint error-and-erasure decoding be achieved for irreducible polynomial remainder codes without incurring high computational overhead from repeated interpolation?
• RQ2 How can existing error-only decoding algorithms be extended to handle erasures while maintaining compatibility with the original implementation?
• RQ3 Is there a GCD-based decoding method for irreducible polynomial remainder codes that reduces computational complexity compared to prior approaches, especially when erasures are present?
• RQ4 Does the proposed second approach offer new decoding capabilities for Reed-Solomon codes, even though it is compatible with existing error-only decoding?
• RQ5 What are the necessary and sufficient conditions under which the GCD-based decoding algorithm correctly recovers the original message polynomial?

Key findings

• The proposed Fixed-Transform Approach I yields a decoding algorithm that is identical to the method in [12] when recovering $a(x)$ via $r(x)/(t(x)\Lambda_\rho(x))$, but introduces a new recovery method using $\hat{Y}(x)$ that is not previously reported.
• Fixed-Transform Approach II, which uses $\tilde{M}_n(x)$ and $Y(x)$ as inputs, is shown to be computationally more efficient than Approach I due to smaller input polynomial degrees.
• The second approach is entirely new even when specialized to Reed-Solomon codes, offering a novel decoding path that maintains compatibility with error-only decoding methods.
• The GCD-based decoding algorithms ensure correct decoding when the error pattern satisfies the condition $\deg E(x) < \deg \Lambda_\rho(x)$, and decoding fails otherwise.
• The algorithms correctly recover $a(x)$ only if $t(x)\Lambda_\rho(x)$ divides $r(x)$ in Approach I or $t(x)$ divides $r(x)$ in Approach II, and if $\deg a(x) < K$.
• The decoding process is fully compatible with existing error-only decoding, enabling seamless integration into systems that already use the error-only variant of the algorithm.