Ordered Statistics Decoding Based on Codeword Neighborhood Structure Learning
IEEE Access · 2026 · 14 · 57475-57489
DOI: 10.1109/access.2026.3682786 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Yuzhou Shang und 2 weiteren Autor:innen. Die Arbeit nennt 26 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Natural Science Foundation of China, Ministry of Industry and Information Technology of the People's Republic of China, in part by Zhejiang Provincial Key Research and Development Program.
Abstract (Originalsprache)
To resolve the computational complexity of ordered statistic decoding (OSD) while preserving its near-maximum likelihood (ML) performance, we propose the codeword neighborhood structure learning-based OSD (CNSL-OSD) decoder. Unlike heuristic termination strategies or black-box neural decoders, our approach establishes a rigorous geometric foundation for the ML decision process by formally defining the codeword neighborhood space (CNS). We analytically derive the minimal subset of neighbor codewords required for precise ML decisions. We prove that the CNS is invariant across the entire codebook, which allows us to train a neural module on the all-zero codeword while guaranteeing robust generalization to arbitrary codewords. By exploiting these intrinsic structural properties of linear block codes, we reformulate the decoding optimization as a geometry-driven ML-consistency learning task. Leveraging this theoretical framework, we incorporate a lightweight, structure-aware neural decision module specifically designed to capture the spatial features of the CNS. This structure-aware design significantly reduces training data requirements while ensuring robust generalization. Furthermore, we develop a sequence-independent search strategy that shifts the computational burden of error pattern (EP) generation from the online decoding phase to an offline pre-computation stage. Experimental results demonstrate that CNSL-OSD achieves near-ML performance with a substantial reduction in average complexity and latency. Specifically, it decreases the required EPs compared to high-order OSD, offering a clear advantage for latency-critical short block code applications.
Förderung
- National Natural Science Foundation of China
- Ministry of Industry and Information Technology of the People's Republic of China
- in part by Zhejiang Provincial Key Research and Development Program
- Fundamental Research Funds for the Central Universities
Zitieren
APA
Shang, Y., Zhang, Z., & Yang, Z. (2026). Ordered Statistics Decoding Based on Codeword Neighborhood Structure Learning. IEEE Access, 14, 57475-57489. https://doi.org/10.1109/access.2026.3682786
Vancouver
Shang Y, Zhang Z, Yang Z. Ordered Statistics Decoding Based on Codeword Neighborhood Structure Learning. IEEE Access. 2026;14:57475-57489. doi:10.1109/access.2026.3682786
BibTeX
@article{Shang2026,
author = {Shang, Yuzhou and Zhang, Zhaoyang and Yang, Zhaohui},
title = {Ordered Statistics Decoding Based on Codeword Neighborhood Structure Learning},
journal = {IEEE Access},
year = {2026},
volume = {14},
pages = {57475--57489},
doi = {10.1109/access.2026.3682786}
}
EUVIMED