LDPC Bit-Flipping Decoding with Inertial Energy Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing LDPC code decoding methods, such as GDBF, often get stuck in local extrema, leading to sub-optimal performance and error correction limitations compared to BP-type decoding methods, which are more resource-intensive.
Innovation Solution
Incorporating an inertial component into the local energy function to reduce the likelihood of bit inversion if a bit has been recently flipped, promoting directional coherence and avoiding oscillations between local extrema, thereby enhancing the decoder's ability to escape local minima.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If GDBF decoding method is used, then computational efficiency is improved, but error correction performance deteriorates due to getting stuck in local extrema
Solution Approach 1:
The patent applies periodic action by implementing a multi-stage decoding process that alternates between GDBF iterations and BP iterations. The decoder performs a first number of GDBF iterations to quickly reduce errors, then switches to BP iterations to achieve optimal performance. This periodic switching combines the computational efficiency of GDBF with the reliability of BP, resolving the contradiction between speed and accuracy.
Solution Approach 2:
The patent changes parameters by dynamically adjusting the decoding strategy based on iteration count and performance metrics. It transitions from using only GDBF (fast but suboptimal) to a hybrid approach where the proportion of BP iterations increases as decoding progresses. This parameter change allows the system to achieve both high computational efficiency in early stages and high reliability in later stages.
2Reliability
If BP decoding method is used, then error correction performance is improved, but resource consumption increases
Solution Approach 1:
The patent applies partial action by using BP decoding for only a portion of the total decoding iterations rather than exclusively. The system performs some GDBF iterations (partial use of fast method) followed by BP iterations (partial use of accurate method). This partial application of each method achieves good error correction performance while consuming fewer resources than full BP decoding.
Solution Approach 2:
The patent segments the decoding process into distinct phases: an initial phase using GDBF for rapid error reduction, followed by a refinement phase using BP for optimal performance. This segmentation allows each method to be applied where it is most effective, reducing overall resource consumption while maintaining high reliability.
3Device complexity
If GDBF decoding method is used, then computational complexity is reduced, but decoding accuracy deteriorates due to local minimum trapping
Solution Approach 1:
The patent introduces BP decoding as an intermediary mechanism that bridges the gap between simple GDBF and optimal decoding. The hybrid approach uses BP iterations as an intermediary step to escape local minima that trap pure GDBF, thereby improving decoding accuracy without proportionally increasing computational complexity.
Solution Approach 2:
The patent maintains continuous useful action by ensuring that both GDBF and BP iterations contribute meaningfully to the decoding process. The transition from GDBF to BP is seamless, with each phase building on the previous phase's results. This continuity ensures that computational resources are always productively utilized, maintaining low complexity while progressively improving accuracy.
Data Source
Figure 1
Figure 2
Figure 3A~3B
AI summary
The present invention relates to a method for decoding LDPC (Low Density Parity Check) codes, of the GDBF (Gradient Descent Bit Flipping) type. At each decoding iteration, as long as the syndrome of the estimated word indicates an error (220,230), a set (F̃) of the least reliable bits of said word is determined (270), such as those whose local energy function (Ẽn) value is less than a threshold value. The local energy value of a bit comprises a first component proportional to the correlation between that bit and a corresponding sample of the observed signal, a second component representing the number of unchecked constraints in which said bit appears, and a third component, called the inertial component, decreasing with the number (ℓn) of iterations performed since the last flip of that bit.The bits of the estimated word belonging to this set are reversed (280), if necessary with a predetermined probability, to provide a new estimated word at the next iteration.