Research On Nonlinear Equalization Method For Optical Communication System With High-Order Modulation Format

With the continuous rise of 8K TV,online games,virtual reality and other Internet services,people’s lives are filled with massive data,which poses a severe challenge to the contemporary communication capacity.Optical communication system,as an important component of modern telecommunication network,is developing towards high speed and large capacity in order to provide better communication services.Coherent optical communication system,which combines high-order modulation formats,coherent detection and digital signal processing technology,has the advantages of flexible modulation formats,high transmission rate and large transmission capacity.It has become an important technology to meet the current capacity demand.However,due to the nonlinear effect of optical fiber,the signal will suffer serious nonlinear damage during transmission,which markedly deteriorates the transmission quality of the signal.Therefore,in order to improve the transmission performance of the system,the research of nonlinear equalization method for coherent optical communication system has become a research hotspot at home and abroad.Focusing on the nonlinear problem in coherent optical communication system with high-order modulation formats,this thesis mainly studies three intelligent optimization nonlinear equalization methods which are respectively optimized Gaussian mixture model algorithm,beetle antennae search-density peak clustering algorithm and K-means-tailored K-nearest neighbors(K-means-TKNN)algorithm,so as to improve the transmission performance of the system.The main research work of this thesis is as follows:(1)In order to alleviate the nonlinear damage of coherent optical communication system,a nonlinear equalization method based on optimized Gaussian mixture model algorithm is proposed.The method adopts an unsupervised neural network(UNN)designed by this thesis to initialize the parameters.UNN has a simple structure and only consists of two layers of neural networks.It can output relatively accurate clustering centers for better determining the initial parameters without introducing training data.Then the initial parameters are provided to the expectation maximization algorithm for iterative optimization to complete the clustering process.The bit error rate and computational complexity performance of the proposed method are studied by the simulation.The results show that,in a single-polarization 16-QAM coherent optical communication system,compared with the maximum likelihood decision method,the proposed method can obtain a nonlinear tolerance gain of 0.84dB under the soft-decision forward error correction threshold with a bit error rate of 2.0 ×10-2.Compared with Gaussian mixture model algorithm,the bit error rate of the proposed method is basically the same,and the computational complexity at high power is reduced by 17.2%.(2)In order to improve the nonlinear tolerance of optical communication system with high-order QAM formats,a nonlinear equalization method based on beetle antennae search-density peak clustering algorithm is proposed.The method adopts the beetle antennae search algorithm to adaptively find the optimal parameter truncation distance,which avoids the subjectivity of manually selecting the parameter.According to this parameter,the local density and relative distance of constellation points are calculated.Then the decision diagram is constructed and the remaining constellation points are allocated to complete clustering process.The bit error rate performance of the proposed method is studied by simulation.The results show that,in a 128-QAM polarization multiplexing optical communication system,compared with the classical density peak clustering algorithm and the maximum likelihood decision method,the proposed method can achieve nonlinear tolerance gains of 0.80dB and 1.46dB respectively under the hard-decision forward error correction threshold with a bit error rate of 3.8 ×10-3.The method can effectively track the distribution of constellation points of high-order QAM signals,and has a good nonlinear equalization effect.(3)In order to alleviate the nonlinear damage of heterodyne coherent detection radio over fiber(ROF)system,a nonlinear equalization method based on K-means-TKNN algorithm is proposed.The method belongs to supervised learning,which introduces training data to improve the classification accuracy of the model.The method adopts multi-level processing of test data and tailoring of training data to reduce the computational complexity,which is helpful to alleviate the complex signal damage from optical fiber and wireless transmission in ROF system.The bit error rate and computational complexity performance of the proposed method are studied by the simulation.The results show that,in a 64-QAM heterodyne coherent detection ROF system,the proposed method can achieve the same bit error rate as K-nearest neighbor(KNN)algorithm,and its computational complexity is reduced by 79.5%compared with KNN algorithm.The proposed method effectively improves the transmission performance of the ROF system.