论文部分内容阅读
为解决差错反向传输神经网络在透明可重构光网络光性能监测中精度不足的问题,提出一种基于优化的径向基函数人工神经网络的光性能监测方案。在该方案中,以信号眼图参数为网络输入,以光信噪比、色散和偏振模色散为网络输出;采用二进制与十进制相结合编码的递阶粒子群方法,用适应度函数引导粒子向小规模和小误差方向运动,进行神经网络的结构与参数自适应优化;分别以不同光信噪比,不同色散和偏振模色散水平仿真信道中传输速率为40 Gb/s差分相移键控仿真信号,进行网络训练和测试,并将测试结果与相同情形下基于差错反向传输法神经网络的光性能监测结果进行比较。结果表明,所提方案在保有人工神经网络方案优点的基础上,有着更好的监测精度。
In order to solve the problem of insufficient accuracy of the reverse transmitted neural network in the optical performance monitoring of transparent reconfigurable optical networks, an optical performance monitoring scheme based on optimized radial basis function artificial neural networks is proposed. In this scheme, signal eye parameters are input to the network, and optical signal-to-noise ratio, dispersion and polarization mode dispersion are output as the network output. The hierarchical particle swarm optimization method combining binary and decimal coding is used to guide the particle orientation Small-scale and small-error direction, and the structure and parameters of the neural network are adaptively optimized. The simulation results of differential phase-shift keying with transmission rates of 40 Gb / s in different optical signal-to-noise ratio, different chromatic dispersion and polarization mode dispersion Signal, network training and testing, and compare the test results with the optical performance monitoring results based on the error feedback neural network in the same situation. The results show that the proposed scheme has better monitoring accuracy based on the advantages of artificial neural network.