.png){kind=icon}
.png){kind=icon}
Return
TS-SCLF: Threshold-Learning-Based SCLF Polar Decoder With Segmented CRC
.png){kind=icon}
PRE
AI
DOI：10.1109/LWC.2025.3627410.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Successive cancellation list flip (SCLF) polar decoding has garnered extensive research attention thanks to its outstanding error correction performance. However, it suffers from high complexity and long worst-case decoding latency, particularly at low signal-to-noise ratios (SNRs). To address these issues, this letter proposes TS-SCLF, an SCLF decoder that employs a dual-threshold strategy and segmented CRC to reduce latency. The proposed dual-threshold strategy dynamically skips redundant flipping attempts through historical failure pattern analysis, while employing Gaussian distribution-inspired path metric thresholds to enable early successive cancellation list (SCL) termination. Furthermore, a deep Q-network (DQN) framework is applied to optimize the threshold parameters, achieving a good balance between complexity and error performance. Experimental results show that for a (1024, 512) polar code, TS-SCLF reduces the average number of decoding attempts by up to 55.3% at low SNRs compared with state-of-the-art counterparts, while preserving error correction performance.
Keywords:
Successive cancelation list flip
polar codes
reinforcement learning
early termination
Journal
I
IF:
5.5
Papers:
663
Citations:
0