Research On Modulation Format Identification Based On Artificial Neuronal Network And Asynchronous Amplitude Histogram Optimized By Genetic Algorithm

End users have ever-growing demands for broadband data and high rate application in modern information society, which has led to the incessant upgrade of the existing networks. With the need of increasing amount of data, the fast development of different technologies such as the single-mode fiber, the single-frequency laser, the Erbium-doped fiber amplifiers(EDFA), wavelength-division multiplexing(WDM), and advanced modulation schemes are pushing transmission capacities very close to the ultimate Shannon capacity limit. Optical networks continue to grow towards high capacity and flexibility, and become heterogeneous to support a wide range of data traffic. It can be foreseen that future optical communication systems are no longer in a fairly static fashion and built to operate within well-defined specifications, the paths and links of dynamic networks may change with temperature, component replacement, aging, fiber plant maintenance and the simple ’plug and play’ optical nodes are needed to allocate resources of the existing optical network. To ensure error-free transmission and quality of service(QoS) agreements, optical performance monitoring(OPM) techniques should provide management, control and optimization services and are essential in dynamic optical networks. OPM monitor important parameters of optical signal, such as optical signal-to-noise ratio(OSNR), modulation formats(MFs) and the key parameters of fiber link such as chromatic dispersion(CD), polarization mode dispersion(PMD) and fiber nonlinear induced penalties. To support the heterogeneous nature such as incorporate multiple modulation formats(MFs) and mixed line rates(MLR)(such as 10/40/100 Gbps), OPM techniques should be equipped with modulation format identification(MFI) capabilities that can recognize the actual modulation format(MF) type of optical signal accurately and OPM devices can be deployed at the intermediate network nodes to apply a specific monitoring.Asynchronous amplitude histograms(AAHs) analysis and artificial neural network(ANN) model are effective methods for OPM and demonstrated good estimation accuracies, especially in modulation format identification(MFI). A cost-effective OPM technique in heterogeneous fiber-optic networks uses an ANN trained with the features extracted from AAHs to recognize six different widely-used modulation formats(MFs), which is the MFI model based on ANN and AAH.In this paper, genetic algorithm(GA) is introduced into the MFI model based on ANN and AAH. The researches cover as follows:First, genetic algorithm(GA) is introduced to optimize the weights and thresholds of ANN. The MFI model based on GA-optimized ANN and AAH could simplify the structure, realize superior overall estimation accuracy and enhance the efficiency of OPM.Second, motivated by the technique known as “sparse sampling”, a method to use genetic algorithm(GA) to select sparse AAHs with equal features of signals as the inputs of the the MFI model based on ANN and AAH is proposed. The MFI model based on ANN and GA-selected sparse AAH could simplify the structure, reduce the computation complexity and enhance the efficiency of OPM.Numerical simulations were performed for six commonly-used MFs at various data rates, including 10 Gbps non-return-to-zero on-off keying(NRZ-OOK), 40 Gbps optical duobinary(ODB), 40 Gbps non-return-to-zero differential phase-shift keying(NRZ-DPSK), 40 Gbps return-to-zero differential quadrature phase-shift keying(RZ-DQPSK), 100 Gbps polarization-multiplexed return-to-zero quadrature phase-shift keying(PDM-RZ-QPSK) and 200 Gbps polarization-multiplexed non-return-to-zero 16 quadrature amplitude modulation(PDM-NRZ-16QAM). Simulations results demonstrate that the proposed MFI model can enhance the efficiency of OPM.