基于并行神经网络的塑料光纤链路自动识别模型

Automatic Identification Model of Plastic Optical Fiber Link Based on Parallel Neural Network

为提升塑料光纤链路识别性能,设计了基于并行神经网络的塑料光纤链路自动识别模型,应用长短期记忆神经网络(LSTM)捕获塑料光纤链路信号的时空依赖特征,应用并行卷积神经网络捕获多样化、抽象、非线性的塑料光纤链路特征信息。经仿真,当批次样本量为30时,模型对10个塑料光纤通信系统的识别准确率分别为96%、95%、96%、95%、99%、96%、96%、97%、96%和95%,平均识别准确率为96.1%,平均识别时间为5.8 s,明显优于骨干光通信网链路识别方案和面向多层次异构信息平台的数据访问链路识别方案。研究成果可为当前塑料光纤链路识别工作提供一定的参考。

and nonlinear characteristics of plastic optical fiber link information.Through simulation,when the batch sample size was 30,the identification accuracy of the model for 10 plastic optical fiber communication systems was 96%,95%,96%,95%,99%,96%,96%,97%,96%and 95%respectively,the average identification accuracy was 96.1%,and the average identification time was 5.8 s,which was obviously superior to the backbone optical communication network link identification scheme and the data access link identification scheme for multi-level heterogeneous information platform.The research results can provide some reference for the current plastic optical fiber link identification.