针对现有基于神经网络的极化码译码算法存在的码长不变码率一旦发生变化就需重新训练与无法充分利用冻结比特位信息的问题，提出一种基于CNN-SENet（ConvolutionalNeural Network-Squeeze and Excitation Networks）的水声信道极化码译码算法。神经网络的输入为双通道输入，通道一为LLR（Log-Likelihood Ratio）信息，通道二为冻结比特位信息，并使用SENet注意力模块对不同通道赋予不同权重，注意不同通道的信息重要程度。仿真结果表明，信号通过水声信道后，在-1～2.5 dB信噪比范围下，对于（8,4）码率的极化码，CNN-SENet译码与SC译码性能相近；对于(8,5)码率的极化码，CNN-SENet译码性能整体高于SC译码，在BER为0.1时，平均可以多提供约0.8 dB的编码增益。
In order to solve the problems of the existing polar code decoding algorithms based on neural networks, the code length is unchanged and the bitrate changes and the frozen bit information cannot be fully utilized, a hydroacoustic channel polarimetric code decoding algorithm based on ConvolutionalNeural Network-Squeeze and Excitation Networks (CNN-SENet) was proposed. The input of the neural network is dual-channel input, channel 1 is Log-Likelihood Ratio (LLR) information, channel 2 is frozen bit information, and SENet attention module is used to assign different weights to different channels to pay attention to the importance of information in different channels. The simulation results show that after the signal passes through the underwater acoustic channel, in the range of -1～2.5 dB signal-to-noise ratio, the performance of CNN-SENet decoding is similar to that of SC decoding for (8,4) bit rate polar codes. For (8,5) bit rate polar codes, the overall decoding performance of CNN-SENet is higher than that of SC decoding, and when the BER is 0.1, it can provide an average of about 0.8 dB more coding gain.
2025,47(16): 117-121 收稿日期：2024-12-2
DOI：10.3404/j.issn.1672-7649.2025.16.018
分类号：U666.7；TN911
基金项目：国家自然科学基金资助项目（62172269）
作者简介：代兵兵(2001-),男,硕士研究生,研究方向为水声信道、深度学习
参考文献：
[1] 翟玉爽, 冯海泓, 李记龙. 极化码在OFDM水声通信中的应用研究[J]. 声学技术, 2021, 40(1): 29-38.
ZHAI Y S, FENG H H, LI J L. Research on the application of polar code in OFDM underwater acoustic communication[J]. Acoustic Technology, 2021, 40(1): 29-38.
[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] 3GPP. Multiplexing and Channel Coding: 3GPP 38.212 V. 15.1. 0 [S]. 2018.
[4] ZHANG Y X, LIU A J, PAN X F, et al. A modified belief propagation polar decoder[J]. IEEE Communications Letters, 2014, 18(7): 1091-1094.
[5] 邱开虎, 黄志亮, 张莜燕, 等. 3×3核矩阵极化码的BP译码算法[J]. 无线电通信技术, 2024, 50(1): 168-172.
QIU K H, HUANG Z L, ZHANG Y Y, et al. BP decoding algorithm for 3×3-core matrix polarization code[J]. Radio Communication Technology, 2024, 50(1): 168-172.
[6] ZHOU L X, CHAN S, ZHANG M X, et al. A fast computing decoder for polar codes with a neural network[J]. ICT Express, 2023, 9(6): 1001-1006.
[7] ZHU H F, CAO Z W, ZHAO Y P, et al. Learning to denoise and decode: a novel residual neural network decoder for polar codes[J]. IEEE Transactions on Vehicular Technology, 2020, 69(8): 8725-8738.
[8] FANG J, BI M, XIAO S, et al. Neural successive cancellation polar decoder with tanh-based modified LLR over FSO turbulence channel[J]. IEEE Photonics Journal, 2020, 12(6): 1-10.
[9] 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.
[10] WANG X M, LI J, CHANG H, HE J L. Optimization design of polar-LDPC concatenated scheme based on deep learning[J]. Computers & Electrical Engineering, 2020, 84: 106636.
[11] 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.
GUO R, RAN F C. Polar code decoding algorithm based on convolutional neural network[J]. Telecommunications Science, 2020, 36(6): 119-124.
[12] 邢莉娟, 李卓, 张泽栋. 水声通信中极化码的应用研究[J]. 电子学报, 2022, 50(9): 2096-2101.
XING L J, LI Z, ZHANG Z D. Research on the application of polar codes in underwater acoustic communication[J]. Acta Electronica Sinica, 2022, 50(9): 2096-2101.
[13] 艾宇慧, 惠俊英, 高静. 水声信道相关均衡器仿真研究[J]. 声学学报, 1999, (6): 589-597.
AI Y H, HUI J Y, GAO J. Simulation study on acoustic channel related equalizer[J]. Journal of Acoustics, 1999, (6): 589-597
[14] HU J, SHEN L, ALBANIE S, et al. Squeeze-and-excitation networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42(8): 2011-2023.
