Polar Code Decoding with Parallel LLR Candidate Decisions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing successive cancellation (SC) decoding method for polar codes in 5G communication systems faces high computational complexity and decoding delay due to the need for calculating and sorting log likelihood ratios (LLRs) for each decoded bit, especially when dealing with large quantities of decoded bits.
Innovation Solution
A decoding method that performs a hard decision on each LLR to obtain an original vector, determines candidate vectors through negation and diagnosis operations, and selects optimal candidate vectors to reduce computational complexity by moving path splitting and error correction to intermediate levels, allowing for parallel decision-making on multiple bits.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If successive cancellation list (SCL) decoding method is used to decode each information bit with path splitting and PM value sorting, then decoding performance is improved, but computational complexity and decoding delay increase significantly
Solution Approach 1:
The patent segments the decoding process into two distinct phases: an intermediate level where path splitting and candidate vector generation occur, and a final level where sorting and selection are performed. This segmentation allows the computationally intensive path splitting operation to be distributed across multiple intermediate nodes rather than concentrated at the final decision level, thereby reducing the overall computational burden while maintaining decoding performance through the preservation of multiple candidate paths.
Solution Approach 2:
The patent performs path splitting and generates candidate vectors in advance at intermediate levels before the final decoding decision. By preliminarily creating and storing candidate vectors at intermediate stages, the system avoids the need to perform exhaustive path exploration and sorting at the final level, thus reducing computational complexity and decoding delay while ensuring that viable candidate paths are already prepared for final selection.
2Measurement precision
If SCL decoding method is used with path splitting for each decoded bit, then decoding accuracy is improved, but decoding delay increases due to sorting operations
Solution Approach 1:
The patent divides the decoding timeline into intermediate processing stages and a final selection stage. Path splitting and candidate vector generation are segmented and performed in parallel at intermediate levels, while the time-consuming sorting operation is confined to the final level where only a limited number of candidate vectors need to be compared. This temporal segmentation reduces decoding delay while preserving decoding accuracy through maintained candidate path diversity.
Solution Approach 2:
The patent performs candidate vector generation and preliminary filtering in advance at intermediate levels before the final decoding decision. By preliminarily preparing candidate vectors and eliminating obviously suboptimal paths at intermediate stages, the system reduces the number of vectors requiring final sorting and comparison, thus decreasing decoding delay while maintaining decoding accuracy through the preservation of promising candidate paths.
3Reliability
If exhaustive expansion method is used for ML decoding, then decoding performance is maximized, but computational complexity becomes prohibitive when M>4
Solution Approach 1:
The patent extracts and removes redundant or suboptimal candidate vectors during the intermediate processing stages, keeping only the most promising candidates for further consideration. By taking out and eliminating clearly inferior paths early in the decoding process, the system significantly reduces the number of vectors that require exhaustive evaluation, thereby lowering computational complexity from exponential to manageable levels while maintaining decoding performance through the preservation of high-quality candidate paths.
Solution Approach 2:
The patent performs path splitting and candidate vector generation for a limited number of intermediate levels rather than exhaustive exploration of all possible paths to the final decision. By applying partial action—generating candidate vectors only up to a certain intermediate depth and then selecting the best candidates—the system achieves near-exhaustive decoding performance with significantly reduced computational complexity, avoiding the prohibitive cost of complete exhaustive expansion.
Data Source
AI summary
A decoding method and apparatus are provided, to improve a degree of parallelism in decoded bit decisions and reduce a decoding delay. The method includes: performing a hard decision on each LLR in an inputted LLR vector having a length of M to obtain a first vector, where M≤N and N is a length of to-be-decoded information; sequentially performing negation of some elements of the first vector to obtain L vectors; and then determining decoding results of the LLR vector based on the L vectors.