An Early-Stopping Mechanism for DSCF Decoding of Polar Codes

TL;DR

This paper introduces an early-stopping mechanism for DSCF decoding of polar codes, significantly reducing average decoding time and variance with minimal error-correction performance loss.

Contribution

It proposes a novel early-stopping metric and modifies DSCF decoding to improve efficiency while maintaining decoding accuracy.

Findings

Reduces average decoding time by up to 22%.

Decreases execution-time variance by up to 45%.

Maintains near-original error-correction performance with only 0.05 dB loss.

Abstract

Polar codes can be decoded with the low-complexity successive-cancellation flip (SCF) algorithm. To improve error-correction performance, the dynamic successive-cancellation flip (DSCF) variant was proposed, where the resulting error-correction performance is similar to that of the successive-cancellation list algorithm with low to moderate list sizes. Regardless of the variant, the SCF algorithm exhibits a variable execution time with a high (worst-case) latency. In this work, we propose an early-stopping metric used to detect codewords that are likely undecodable such that the decoder can be stopped at earlier stages for those codewords. We then propose a modified version of the DSCF algorithm that integrates our early-stopping metric that exploits the specific properties of DSCF. Compared to the original DSCF algorithm, in the region of interest for wireless communications, simulation results show that our proposed modifications can lead to reductions of 22% to the average execution time and of 45% to the execution-time variance at the cost of a minor error-correction loss of approximately 0.05 dB.