Low-Complexity Soft-Aided Error-and-Erasure Decoding for Generalized Product Codes

2026-07-15

AI Summary

This work addresses the high error floor and limited performance of high-rate generalized product codes under conventional hard-decision decoding. To overcome these limitations, the paper proposes a Reliability-Directed Syndrome-domain hard-message-passing iterative decoder (RDRSD) that incorporates a dynamic reliability scoring mechanism. This approach achieves significant decoding gains while maintaining hardware complexity comparable to that of iterative bounded-distance decoding (iBDD). Furthermore, by integrating a soft-information-aided post-processing strategy, the proposed method effectively suppresses the error floor. Experimental results demonstrate that the RDRSD scheme yields approximately 1 dB coding gain over iBDD, with particularly pronounced error-floor reduction in scenarios employing component codes with small error-correction capability.

Abstract

Generalized product codes (GPCs) combine excellent high-rate performance with low-complexity hardware implementations. We propose the refined dynamic reliability score decoder (\acs{RDRSD}), a hard-message-passing iterative error-and-erasure decoder that uses dynamic reliability scores. Its syndrome-domain implementation has complexity comparable to hard-decision \ac{iBDD}. Across various GPCs and decoding configurations, \acs{RDRSD} provides different complexity--performance trade-offs and achieves approximately \SI{1}{dB} coding gain over \ac{iBDD}. For GPCs with component codes of small error-correcting capability, we also analyze the error floor and propose a soft-aided post-processing step that significantly lowers it.

Problem

• Generalized Product Codes
• Error-and-Erasure Decoding
• Low Complexity
• Error Floor
• Soft-Aided Decoding

Innovation

• Generalized Product Codes
• Error-and-Erasure Decoding
• Dynamic Reliability Score
• Low-Complexity Decoding
• Soft-Aided Post-Processing