A Normalized Offset Minimum Sum Decoding Method for LDPC Codes Based on Grouped Neural Network Optimization

By converting the Tanner graph of LDPC codes into a grouped neural network and dynamically optimizing the normalization factor and offset factor, the error accumulation problem of LDPC decoding algorithms in 6G scenarios is solved, achieving efficient and robust decoding performance, applicable to LDPC codes of different code lengths and rates.

CN122372006APending Publication Date: 2026-07-10CHONGQING UNIV OF POSTS & TELECOMM
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
CHONGQING UNIV OF POSTS & TELECOMM
Filing Date
2026-04-09
Publication Date
2026-07-10

AI Technical Summary

Technical Problem

Existing LDPC decoding algorithms suffer from error accumulation issues due to parameter fixation in 6G scenarios, affecting decoding reliability and efficiency, especially with a high frame error rate in ultra-reliable and ultra-low latency communication scenarios.

Method used

By converting the Tanner graph of the LDPC code into a grouped neural network structure, the normalization factor and offset factor are optimized using the neural network, and the parameters are dynamically adjusted to adapt to different iteration rounds and edge types. Combined with the Adam optimizer to optimize the factors, fast decoding is achieved.

Benefits of technology

It significantly reduces error accumulation, improves decoding convergence speed and accuracy, reduces computational complexity and storage overhead, has a wider range of applications, and possesses higher robustness and decoding efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122372006A_ABST
    Figure CN122372006A_ABST
Patent Text Reader

Abstract

This invention relates to a Normalized Minimum Offset Sum (NOMS) decoding method for LDPC codes based on optimized grouped neural networks, belonging to the field of wireless communication channel coding technology. This method utilizes the topological relationship between the LDPC code base graph and the Tanner graph to transform the Tanner graph into a non-fully connected grouped neural network. It groups edges based on the base graph type and achieves parameter sharing using edge bundles. A greedy hierarchical training algorithm is used to optimize the normalization factor and offset factor round by round, obtaining dynamically adaptive parameters that adapt to each iteration stage. The optimized factors are substituted into the decoding update formula to replace the traditional fixed factors and complete the normalized minimum offset sum decoding. This invention can suppress error amplification during iteration, avoid error leveling and decoding deadlock, improve decoding accuracy while reducing computational complexity and parameter size, exhibit strong generalization ability, and can adapt to LDPC codes with different code lengths and rates under the same base graph. It also features low latency, high robustness, and engineering practicality.
Need to check novelty before this filing date? Find Prior Art