基于CNN的塑料光纤信道解码技术研究  

Research on Plastic Optical Fibre Channel Decoding Technology Based on CNN

作　　者：魏赫阳 WEI He-yang(Eurasian International School,Henan University,Kaifeng 475001,China)

机构地区：[1]河南大学欧亚国际学院,河南开封475001

出　　处：《塑料科技》2020年第6期69-73,共5页Plastics Science and Technology

摘　　要：为降低塑料光纤中放大自发发射噪声对解码信号的影响,应用卷积神经网络实现塑料光纤信道解码器,并设计三个TInception类型网络。将TanH非线性激活函数应用于解码器网络第一层,将设计的TInception2应用于解码器网络第二层,可提升训练效率,便于从塑料光纤信号中捕获多样且抽象的信号特征,增强模型的特征表达能力。经仿真,本塑料光纤信道解码器在非对称X^2噪声信道、非对称高斯噪声信道、加性高斯白噪声信道上,对于硬判决和软判决,Q因子为5.75时,信号误码率分别为0.08272、0.00064、0.00726、0.00123、0.00627和0.00134,均低于传统Viterbi光纤信道解码器和反馈神经网络解码器。In order to reduce the influence of amplified spontaneous emission noise on the decoded signal in plastic optical fiber,convolutional neural network is used to realize the decoder of plastic optical fiber channel,and three TInception networks are designed.TanH nonlinear activation function is applied to the first layer of decoder network,and designed TInception2 is applied to the second layer of decoder network,which can improve the training efficiency,facilitate the acquisition of diverse and abstract signal features from plastic optical fiber signals,and enhance the feature expression ability of the model.Through simulation,for hard decision and soft decision,when Q factor was 5.75,the BER of the decoder was 0.08272,0.00064,0.00726,0.00123,0.00627 and 0.00134,which were lower than that of the traditional Viterbi decoder and RNN decoder.

关 键 词：塑料光纤 卷积神经网络 解码器 放大自发发射噪声 

分 类 号：TQ342[化学工程—化纤工业]