基于深度学习的神经归一化最小和LDPC长码译码  

LDPC long code decoding with neural normalized min-sum based on deep learning

作　　者：贾迪 严伟[1] 姚赛杰 张权 刘亚欢 Jia Di;Yan Wei;Yao Saijie;Zhang Quan;Liu Yahuan(School of Software and Microelectronics,Peking University,Beijing 102600,China;Motorcomm Co.,Ltd.,Shanghai 201210,China)

机构地区：[1]北京大学软件与微电子学院,北京102600 [2]裕太微电子股份有限公司,上海201210

出　　处：《电子技术应用》2024年第12期7-12,共6页Application of Electronic Technique

基　　金：四川省科技厅重点研发计划(2023YFG0120)。

摘　　要：LDPC码是一种应用广泛的高性能纠错码,近年来基于深度学习和神经网络的LDPC译码成为研究热点。基于CCSDS标准的(512,256)LDPC码,首先研究了传统的SP、MS、NMS、OMS的译码算法,为神经网络的构建奠定基础。然后研究基于数据驱动(DD)的译码方法,即采用大量信息及其经编码、调制、加噪的LDPC码作为训练数据在多层感知层(MLP)神经网络中进行训练。为解决数据驱动方法误码率高的问题,又提出了将NMS算法映射到神经网络结构的神经归一化最小和(NNMS)译码,取得了比NMS更优秀的误码性能,信道信噪比为3.5 dB时误码率下降85.19%。最后研究了提升NNMS网络的SNR泛化能力的改进训练方法。LDPC code is a widely-used high-performance error correction code.In recent years,LDPC decoding based on deep learning and neural networks becomes a research hotspot.Based on the(512,256)LDPC code of the CCSDS standard,this paper firstly studies the traditional decoding algorithms of SP,MS,NMS,and OMS,laying a foundation for the construction of neural networks.Then a data-driven(DD)decoding method is studied which adopts the information with its encoded,modulated and noise-added LDPC code as the training data within a Multi-layer Perceptron(MLP)neural network.In order to solve the problem of high bit error rate(BER)in data-driven method,the Neural Normalized Min-sum(NNMS)decoding in which the NMS algo-rithm is mapped to the neural network structure is proposed,achieving more excellent BER performance than that of NMS.The BER declines by 85.19%when channel SNR equals to 3.5 dB.Finally,improved training methods to enhance the SNR generaliza-tion ability of the NNMS network is studied.

关 键 词：LDPC 深度学习 神经网络 

分 类 号：TN911.22[电子电信—通信与信息系统]