面向弹性光网络中间节点的嵌入式光性能监测系统  

Embedded Optical Performance Monitoring System for Intermediate Nodes in Elastic Optical Networks

作　　者：周唐磊 曹领国 毕岩峰 王志国 许恒迎[1,2] 白成林[1,2] ZHOU Tanglei;CAO Lingguo;BI Yanfeng;WANG Zhiguo;XU Hengying;BAI Chenglin(School of Physics Science and Information Engineering,Liaocheng University,Liaocheng 252059,China;Shandong Provincial Key Laboratory of Optical Communication Science and Technology,Liaocheng University,Liaocheng 252059,China)

机构地区：[1]聊城大学物理科学与信息工程学院,山东聊城252059 [2]聊城大学山东省光通信重点实验室,山东聊城252059

出　　处：《聊城大学学报（自然科学版）》2022年第4期52-58,64,共8页Journal of Liaocheng University:Natural Science Edition

基　　金：国家自然科学基金项目(61501213,62101229);山东省自然科学基金项目(ZR2020MF012,ZR2020QF005);聊城大学博士科研启动基金项目(318051834,318051835)资助。

摘　　要：针对目前卷积神经网络在弹性光网络中间节点监测时硬件成本高,占用内存大的问题,提出一种基于二值卷积神经网络与异步延时抽头采样技术的嵌入式光性能监测系统。结果表明,对于16QAM、32QAM、64QAM信号,调制格式识别准确率均为100%,光信噪比识别准确率分别为99.1%、98.5%和97.4%。这些结果证明了在嵌入式平台上将异步延时抽头采样技术和二值卷积神经网络用于中间节点进行链路监测的可行性。此外,本系统的内存占用减少为浮点型卷积神经网络的1/4,识别时间缩短为它的1/3,更适合部署于资源有限的弹性光网络中间节点。In order to solve the problems of high hardware cost and large memory consumption of convolutional neural network in the optical performance monitoring for the intermediate nodes of elastic optical network,we have proposed an embedded optical performance monitoring system based on binary convolutional neural network and asynchronous delay tap sampling technology.For 16QAM,32QAM and 64QAM signals,the results show that the identification accuracy of modulation format is 100%,and the accuracies of optical signal-to-noise ratio are 99.1%,98.5%and 97.4%,respectively.These results prove that the asynchronous delay tap sampling technology and binary convolutional neural network can be used for link monitoring of intermediate nodes on embedded platform.In addition,compared with float valued convolutional neural network,the memory usage of our system is reduced to 1/4 and the identification time of this system can be shortened to 1/3.Therefore,it is more suitable for the deployment in the intermediate node of elastic optical network with limited resources.

关 键 词：弹性光网络 二值卷积神经网络 异步延时抽头采样 相干光通信 嵌入式系统 

分 类 号：K826.1[历史地理—历史学]