基于遗传算法优化神经元网络和异步幅度图的调制格式识别研究

发布时间：2018-05-22 15:41

本文选题：光性能监测 + 调制格式识别　；参考：《暨南大学》2016年硕士论文

【摘要】：在现代信息社会中,终端用户对宽带数据和高速应用的需求正不断增长,这驱动着通信网络不断升级。为了不断增加数据传输量,单模光纤(single-mode fiber)、窄线宽激光器(single-frequency laser)、掺铒光纤放大器(EDFA)、波分复用技术(WDM)和高级的调制方案等技术正快速发展,这使得光通信网络的传输容量已非常接近于香农定理下的传输容量限制。光通信网络将继续增大容量和灵活性,并采用异质传输来支持广泛的数据传输。可以预见,未来的光通信系统将不再是一个操作完全遵循既成规范的相对静态的网络系统,动态的链路结构将随着温度,元件更换、老化,光纤维护而改变,并且需要能够实现“即插即用”的光节点更好地分配光网络资。为了保证无误差传输和信号服务质量(quality of service,QoS),光性能监测(OPM)技术中能提供信号管理、控制和优化的服务,在动态光网络中至关重要。OPM监测光信噪比(optical signal-to-noise ratio,OSNR)、调制格式(modulation formats,MFs)和光纤链路的重要参数如色散(chromatic dispersion,CD)、偏振模色散(polarization mode dispersion,PMD)、光纤非线性干扰等。未来光通信网络将存在多个调制格式(MFs)和线路速率(MLR)(如10/40/100 Gbps),为了支持这种异质性,OPM技术应具备调制格式识别(modulation format identification,MFI)的能力,可以准确识别光信号的调制格式并且可接入网络节点做特定监测。异步幅度图(Asynchronous amplitude histograms,AAHs)分析和人工神经网络(artificial neural network,ANN)模型是实现OPM技术有效的方法,并且表现出良好的估算精度,尤其是在调制格式识别(MFI)技术中。在异质光通信网络中,利用异步幅度图作为神经元网络的特征输入可以有效地实现调制格式识别,这是一种基于神经元网络和异步幅度图的调制格式识别模型。在本文中,遗传算法(GA)被引入到基于神经元网络和异步幅度图的调制格式识别模型中。该研究涵盖如下:首先,使用遗传算法优化用于识别调制格式的神经元网络的权值和阈值。基于遗传算法优化的神经元网络和异步幅度图的调制格式识别模型可以实现更简化的结构、更高的识别精度来提高光性能监测技术的效率。再者,基于“稀疏采样”技术的启发,本文提出了一种通过遗传算法筛选出等价于完整异步幅度图的稀疏异步幅度图的方法,将稀疏的异步幅度图作为基于神经元网络和异步幅度图的调制格式识别模型的特征输入。基于神经元网络和遗传算法筛选出的稀疏异步幅度图的调制格式识别模型可以通过更简化的模型结构、更低的运算复杂度来提高光性能监测技术的效率。我们对不同损伤下的六种常用的调制格式:10 Gbps NRZ-OOK,40 Gbps ODB,40 Gbps NRZ-DPSK,40 Gbps RZ-DQPSK,100 Gbps PDM-RZ-QPSK和200 Gbps PDM-NRZ-16QAM做了大量的仿真实验。实验结果证明这两种新的调制格式识别模型能提高光性能监测技术的效率。
[Abstract]:In the modern information society, the demand of end users for broadband data and high-speed applications is growing, which drives the continuous upgrading of communication networks. In order to increase data transmission, single-mode optical fiber single-mode laser single-frequency laser with narrow linewidth, erbium-doped fiber amplifier EDFAA, wavelength division multiplexing (WDM) and advanced modulation schemes are developing rapidly. This makes the transmission capacity of optical communication network very close to the limit of transmission capacity under Shannon theorem. Optical communication networks will continue to increase capacity, flexibility and heterogeneous transmission to support a wide range of data transmission. It can be predicted that the future optical communication system will no longer be a relatively static network system that operates completely in accordance with the established specifications, and the dynamic link structure will change with temperature, component replacement, aging, and optical fiber maintenance. And the optical nodes that can realize plug-and-play are needed to allocate the network capital better. In order to ensure error-free transmission and quality of service quality of QoS (quality of service), optical performance monitoring (OPMN) technology can provide signal management, control and optimization services. In dynamic optical networks, it is very important to monitor optical signal-to-noise optical signal to noise ratio (SNR), modulation formats (MFs) and important parameters of fiber links such as dispersion chromatic dispersion, polarization mode dispersion (PMD), optical fiber nonlinear interference and so on. In the future, optical communication networks will have multiple modulation formats (MFs) and line rate (such as 10 / 40 / 100 Gbpss). In order to support this heterogeneity, OPM technology should have the capability of modulation format recognition and modulation format identification. The modulation format of the optical signal can be accurately identified and can be accessed to the network node for specific monitoring. Asynchronous amplitude histogramsm analysis and artificial neural network Ann model are effective methods to realize OPM technology, and show good estimation accuracy, especially in modulation format recognition. In heterogeneous optical communication networks, modulation format recognition can be realized effectively by using asynchronous amplitude graph as the characteristic input of neural network, which is a modulation format recognition model based on neural network and asynchronous amplitude diagram. In this paper, genetic algorithm (GA) is introduced into the modulation format recognition model based on neural networks and asynchronous amplitude diagrams. The research covers the following aspects: firstly, genetic algorithm is used to optimize the weights and thresholds of neural networks to identify modulation formats. The neural network based on genetic algorithm and the modulation pattern recognition model based on asynchronous amplitude diagram can achieve more simplified structure and higher recognition accuracy to improve the efficiency of optical performance monitoring technology. Furthermore, based on the heuristic of "sparse sampling" technique, this paper proposes a method of screening sparse asynchronous amplitude graphs equivalent to complete asynchronous amplitude graphs by genetic algorithm. The sparse asynchronous amplitude graph is used as the feature input of the modulation format recognition model based on neural network and asynchronous amplitude diagram. The modulation format recognition model based on neural network and genetic algorithm can improve the efficiency of optical performance monitoring technology by simplified model structure and lower computational complexity. We have done a lot of simulation experiments on six common modulation formats: 10 Gbps NRZ-OKO 40 Gbps ODBN 40 Gbps NRZ-DPSKN 40 Gbps RZ-DQPSK Gbps PDM-RZ-QPSK and 200 Gbps PDM-NRZ-16QAM under different damage conditions. Experimental results show that the two new modulation pattern recognition models can improve the efficiency of optical performance monitoring technology.
【学位授予单位】：暨南大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TN929.1

【相似文献】