Neural Successive Cancellation Flip Decoding of Polar Codes

• Published in: Journal of signal processing systems Vol. 93; no. 6; pp. 631 - 642
• Main Authors: Doan, Nghia, Hashemi, Seyyed Ali, Ercan, Furkan, Tonnellier, Thibaud, Gross, Warren J.
• Format: Journal Article
• Language: English
• Published: New York Springer US 01-06-2021; Springer Nature B.V
• Subjects: Algorithms; Approximation; Cancellation; Circuits and Systems; Complexity; Computation; Computer Imaging; Decoding; Electrical Engineering; Engineering; Error correction; Image Processing and Computer Vision; Machine learning; Mathematical analysis; Measurement; Parameters; Pattern Recognition; Pattern Recognition and Graphics; Performance degradation; Perturbation; Signal to noise ratio; Signal,Image and Speech Processing; Vision; Deep learning; SC flip; 5G; Polar codes
• ISSN: 1939-8018; 1939-8115
• DOI: 10.1007/s11265-020-01599-y

Dynamic successive cancellation flip (DSCF) decoding of polar codes is a powerful algorithm that can achieve the error correction performance of successive cancellation list (SCL) decoding, with an average complexity that is close to that of successive cancellation (SC) decoding at practical signal-to-noise ratio (SNR) regimes. However, DSCF decoding requires costly transcendental computations to calculate a bit-flipping metric, which adversely affect its implementation complexity. In this paper, we first show that a direct application of common approximation schemes on the conventional DSCF decoding results in a significant error-correction performance loss. We then introduce an additive perturbation parameter and propose an approximation scheme which completely removes the need to perform transcendental computations in DSCF decoding. Machine learning (ML) techniques are then utilized to optimize the perturbation parameter of the proposed scheme. Furthermore, a quantization scheme is developed to enable efficient hardware implementation. Simulation results show that when compared with DSCF decoding, the proposed decoder with quantization scheme only experiences a negligible error-correction performance degradation of less that 0.08 dB at a target frame-error-rate (FER) of 10
− 4
, for a polar code of length 512 with 256 information bits. In addition, the bit-flipping metric computation of the proposed decoder reduces up to around 31
%
of the number of additions used by the bit-flipping metric computation of DSCF decoding, without any need to perform costly transcendental computations and multiplications.