The vigorous development of applications such as big data,cloud computing,mobile Internet,and high-definition video has made Internet technology infiltrating into every aspect of our lives.The surge in network traffic and the number of network users has brought unprecedented challenges to communication technology,which have prompted the development of communication system in the direction of higher speed,larger capacity,and more security and reliability.Many researchers are also exploring different methods to increase the capacity of optical fiber communication system from various aspects.High-order modulation formats can effectively improve the spectral efficiency of the system,and have been widely used in optical fiber communication system.However,there is still a gap between the standard higher-order modulation formats and the Shannon limit.Constellation shaping technology can improve the matching degree of the modulation format and the transmission channel by adjusting the probability or geometric distribution of constellation points,which has become an effective means to approach the Shannon limit and improve the transmission performance of optical fiber communication system.On the basis of studying the architecture and key technologies of optical fiber communication system,this dissertation focuses on the Gray optimization scheme of non-uniform 12QAM signal based on Huffman coding,the three-dimension 14QAM constellation shaping scheme based on regular tetrahedron,the nonlinear equalization scheme of the probabilistic shaping signal based on optimized Gaussian mixture clustering algorithm,and the nonlinear equalization scheme of the probabilistic shaping signal based on constellation segmentation and support vector machine algorithm,aiming to further improve the transmission performance of the optical fiber communication system.The main research work and innovation points of the dissertation are as follows:1.Gray optimization scheme for non-uniform 12QAM signal based on Huffman codingThe traditional non-uniform 12QAM signal based on Huffman coding is without Gray mapping.Compared with the signal with Gray mapping,the signal without Gray mapping will have higher bit error rate under the same symbol error rate.To solve the problem,a Gray optimization scheme for non-uniform 12QAM signal based on Huffman coding is proposed.A single-carrier dual-polarization coherent optical transmission system is built to verify the proposed scheme.The simulation results show that under the HD-FEC decision threshold,compared with the non-Gray mapping 12QAM,the Gray mapping 12QAM obtains about 0.7 dB OSNR gain.2.Three-dimension 14QAM constellation shaping scheme based on regular tetrahedronIn order to further improve the constellation shaping gain,a threedimension 14QAM constellation shaping scheme based on regular tetrahedron is proposed.The scheme uses Huffman coding to design nonuniform 14QAM signal generation scheme,uses an interleaver to generate the non-uniform 14QAM signal,and maps the 14QAM signal to a threedimension constellation diagram based on a regular tetrahedron.A singlecarrier dual-polarization coherent optical transmission system and a wavelength-division multiplexing dual-polarization coherent optical transmission system are built to verify the scheme.Compared with the traditional 3D-16QAM,the proposed scheme can obtain an OSNR gain of 1.1 dB in the single-carrier dual-polarization coherent optical transmission system,and an OSNR gain of 1.8 dB in the wavelength-division multiplexed dual-polarization coherent optical transmission system.The proposed 3D-14QAM combines the advantages of probabilistic shaping and geometric shaping,and extends the constellation design to the threedimension space,providing a broad space for the optimization of the geometric positions and probability distributions of constellation points.3.Nonlinear equalization scheme for probabilistic shaping signal based on optimized Gaussian mixture clustering algorithmIn order to further reduce the influence of nonlinear effects on the optical fiber communication system,a nonlinear equalization scheme for probabilistic shaping signal based on optimized Gaussian mixture clustering algorithm is proposed.Due to the non-uniform distribution of the probabilistic shaping signal,the Gaussian mixture model cannot converge correctly when the traditional Gaussian mixture clustering algorithm is used as the nonlinear equalization algorithm for probabilistic shaping signal.The proposed scheme uses the coordinates of the standard constellation points of PS-64QAM as the initial mean of each Gaussian mixture components,which improves the clustering effect of the Gaussian mixture clustering algorithm on non-uniform PS-64QAM signal.The proposed scheme is experimentally verified using a single-carrier dualpolarization PS-64QAM coherent optical communication system.Compared with the traditional Gaussian mixture clustering algorithm,the optimized K-means algorithm and the direct decision,the dynamic range of the launched optical power of the optimized Gaussian mixture clustering algorithm is improved by 0.42 dB,0.7 dB and 0.7 dB,respectively.4.Nonlinear equalization scheme for probabilistic shaping signal based on constellation segmentation and support vector machineDue to the non-uniform distribution of the probabilistic shaping signal,when the support vector machine is used as the nonlinear equalization algorithm of the probabilistic shaping signal,the hyperplane will move to the direction of the minority points as much as possible.This will cause many minority class points to be wrongly judged to other classes.In order to solve the problem,a nonlinear equalization scheme based on constellation segmentation and SVM algorithm is proposed for probabilistic shaping signal.In this scheme,a SVM is used to divide the constellation properly,and then the M-ary SVM algorithm is used as the nonlinear equalization algorithm of the inner and outer constellation sets respectively.The proposed scheme is verified by the single-carrier dualpolarization PS-64QAM coherent optical communication experimental platform.Compared with the M-ary SVM algorithm,the proposed scheme can obtain a 0.6 dB improvement in the dynamic range of launched optical power,and at the optimal launched optical power of-3 dBm,the proposed scheme can obtain a Q factor improvement of 0.5 dB.In addition,the complexity of the proposed scheme is also effectively reduced compared with the traditional M-ary SVM scheme. |