基于卷积神经网络的极化码译码算法

doi:10.11959/j.issn.1000-0801.2020130

摘要/Abstract

摘要：

针对现有基于神经网络的极化码译码算法仅能对短码进行译码的问题（码长 N≤64），提出针对极化码长码（N≥512）的卷积神经网络译码算法，利用神经网络epoch和batch参数来调控神经网络的输入，而不是固定从数据样本中抽取一定比例用作训练集和测试集，解决了由码字过长造成的数据爆炸问题。此外，研究了batch和epoch参数对卷积神经网络译码性能的影响及神经网络使用不同激活函数的性能差异。仿真结果表明，与传统SCL（successive cancellation list，串行抵消列表）译码算法比较，卷积神经网络在低信噪比下取得略优于SCL（L=2）的性能，在高信噪比下取得与SCL（L=2）算法相近的性能，并且训练数据集越大，神经网络译码性能越好。

关键词: 深度学习, 神经网络, 信道译码, 极化码译码

Abstract:

In order to solve the problem that the existing Polar code decoding algorithm based on neural network can only decode short codewords (codewords length N≤64),a new decoding algorithm using convolution neural network for long codewords (N≥512) was put forward.Instead of using a fixed drawn from the data sample proportion as training set and test set,the neural network parameters epoch and batch were used to control the neural network input,and the data explosion problem caused by the long codewords was solved.In addition,the influence of batch and epoch parameters on the decoding performance of convolution neural network was explored and the performance difference of neural network using different activation functions were investigated.Simulation results show that with the traditional SCL (successive cancellation list,L=2) decoding algorithm,convolution neural network in low signal-to-noise ratio on the performance is better than that of SCL (L=2),in high signal-to-noise ratio and SCL (L=2) algorithm of similar performance,and the larger the training data set,the better the decoding performance of neural network.

Key words: deep learning, neural network, channel decoding, Polar code decoding

中图分类号:

TN911

郭锐,冉凡春. 基于卷积神经网络的极化码译码算法[J]. 电信科学, 2020, 36(6): 119-124.

Rui GUO,Fanchun RAN. Polar codes decoding algorithm based on convolutional neural network[J]. Telecommunications Science, 2020, 36(6): 119-124.

图/表 11

图1

表1

激活函数"

激活函数	表达式	激活函数	表达式
linear	x	Sigmoid	$\frac{1}{e^{x} + 1}$
ReLU	max（0,x）	tanh	$\frac{e^{2 x} - 1}{e^{2 x} + 1}$

表1

图2

表2

图3

图4

图5

图6

图7

图8

图9

参考文献 19

[1]	BOCCARDI F , HEATH R W , LOZANO A ,et al. Five disruptive technology directions for 5G[J]. IEEE Communications Magazine, 2014,52(2): 74-80.
[2]	ARIKAN E . Channel polarization:a method for constructing capacity-achieving codes for symmetric binary-input memoryless channels[J]. IEEE Transactions on Information Theory, 2009,55(7): 3051-3073.
[3]	RUMELHART D E , HINTON G E , WILLIAMS R J . Learning representations by back-propagating errors[J]. Nature, 1986,323(6088): 533-536.
[4]	CAMMERER S , GRUBER T , HOYDIS J ,et al. Scaling deep learning-based decoding of Polar codes via partitioning[C]// Proceedings of GLOBECOM 2017-2017 IEEE Global Communications Conference. Piscataway:IEEE Press, 2017: 1-6.
[5]	NACHMANI E , MARCIANO E , BURSHTEIN D . RNN decoding of linear block codes[J]. 2017.
[6]	SEO J , LEE J , KIM K . Decoding of Polar code by using deep feed-forward neural networks[C]// Proceedings of 2018 International Conference on Computing,Networking and Communications (ICNC). Piscataway:IEEE Press, 2018.
[7]	GRUBER T , CAMMERER S , HOYDIS J ,et al. On deep learning-based channel decoding[C]// Proceedings of 2017 51st Annual Conference on Information Sciences and Systems (CISS).[S.l.:s.n]. 2017: 1-6.
[8]	WANG X , TAN Z H . Development of channel coding technology under wireless channel[J]. Journal of Beijing Jiaotong University, 2004,28(3): 82-87.
[9]	SHANNON C E . A mathematical theory of communication[J]. Bell Labs Technical Journal, 1948,27(4): 379-423.
[10]	MORI R , TANAKA T . Performance and construction of Polar codes on symmetric binary-input memory less channels[C]// Proceedings of IEEE International Symposium on Information Theory. Piscataway:IEEE Press, 2009.
[11]	NIU K , LIN J R , CHEN K . List successive cancellation decoding of Polar codes[J]. Electronics Letters, 2012,48(9): 500-501.
[12]	BALATSOUKAS-STIMMING A , PARIZI M B , BURG A . On metric sorting for successive cancellation list decoding of Polar codes[C]// Proceedings of IEEE International Symposium on Circuits ＆ Systems. Piscataway:IEEE Press, 2015.
[13]	LI Y D , HAO Z B , LEI H . Research review of convolutional neural networks[J]. Computer Applications, 2016,36(9): 2508-2515.
[14]	贾川民, 赵政辉, 王苫社 ,等. 基于神经网络的图像视频编码[J]. 电信科学, 2019,35(5): 38-48.
	JIA C M , ZHAO Z H , WANG J S ,et al. Image and video coding based on neural network[J]. Telecommunications Science, 2019,35(5): 38-48.
[15]	苏赋, 吕沁, 罗仁泽 . 基于深度学习的图像分类研究综述[J]. 电信科学, 2019,35(11): 58-74.
	SU F , LV Q , LUO R Z . Review of image classification based on deep learning[J]. Telecommunications Science, 2019,35(11): 58-74.
[16]	SHELKE S , APTE S . A novel multi-feature multi-classifier scheme for unconstrained handwritten Devanagari character recognition[C]// Proceedings of International Conference on Frontiers in Handwriting Recognitions. Piscataway:IEEE Press, 2010.
[17]	KOLEN J F , KREMER S C . Gradient flow in recurrent nets:the difficulty of learning longterm dependencies[C]// A field guide to dynamical recurrent networks. Piscataway:IEEE Press, 2001: 237-243.
[18]	FANG X N . Principles of four deep learning methods from supervised learning to intensive learning[J]. Robotics Industry, 2017(4): 50-55.
[19]	LYU W , ZHANG Z , JIAO C ,et al. Performance evaluation of channel decoding with deep neural networks[C]// Proceedings of 2018 IEEE International Conference on Communications (ICC). Piscataway:IEEE Press, 2018: 1-6.

隐藏层类型	卷积核大小/步长	输入尺寸
卷积层一	3×1/1	N×1×1
池化层一	2×1/2	N×1×8
卷积层二	3×1/1	N/2×1×8
池化层二	2×1/2	N/2×1×16
卷积层三	3×1/1	N/4×1×16
池化层三	2×1/2	N/4×1×32
卷积层四	N/8×1/1	N/8×1×32
squeeze	降低维度	1×1×K
Sigmoid	分类输出	1×K