Study Of Dumultiplexing Technology Based On ICA In Mode Group Diversity Multiplexing System

Mode group diversity multiplexing (MGDM) technology aims at creating independent, parallel communication channels over a single GI-MMF using several groups of modes, and each mode group can be considered as transporting independently. It is a key technology to realize the efficiency of transport for multi-type operations simultaneously, with the capacity improvement of multi-mode fiber significantly. However, each output signal of the system is the weighted sum of all source signals because of the modes mixing effect in the MMF, thus the crosstalk between the channels appears. How to eliminate the crosstalk and recover the source signals is the chief problem to be resolved, which is also the main task in this thesis.Based on the principles and theories of MGDM technology, a 2×2 MGDM experimental system which realizes the 1km distance transmission efficiently is designed. The demultiplexing method in MGDM system based on the independent component analysis (ICA) is built and developed specially centering on the crosstalk cancellation between channels. The problem of demultiplexing method with noise existence is mainly investigated. And considering the time-domain and frequency-domain distribution characteristics of signal and noise, time and frequency spatial double threshold-based denoising technical based on secondary ICA is proposed. Aiming at the characteristics of MGDM system, order identification based on the correlation coefficient matrix is used to solve the problem of indetermination property about the separated signals of ICA algorithm. The works in this paper are as follows:Firstly, background and experiment research of MGDM system is introduced. Mode dispersion which is the reason for using MGDM technology, and the selective mode excitation which is the basis of multiplexing and demultiplexing, also the mode group which is the unit of information transmitted in MGDM system are fully introduced; the basic theory, the realization and the system modal of MGDM technology are systematic studied, the influence of noise in the MGDM system is analsysed; also, a 2×2 MGDM experimental system is designed to realize the two signals transmitting simultaneously with the output performance being studied using spatial selective detection model.Secondly, basic principle of ICA is introduced. The basic modal of ICA, the method for data preprocessing, the objective function and optimization algorithm of ICA algorithm are described in detail. The emphasis is on the FastICA algorithm based on maximum negentropy. The feasibility of ICA used for signal separation in MGDM system is discussed.Thirdly, investigations on the demultiplexing method using ICA are carried out with the noise existence. After the signification of ICA used for the description about the signal character is introduced, the characteristic extraction model is constructed based on the secondary ICA for the signals with noise existing. Different characteristics with time domain and frequency domain are analyzed; the frequency threshold-based denoising method, also called low-pass filter method, is extracted by analyzing the distribution difference of frequency domain between signal and noise. By means of the time domain statistical distribution difference after ICA transformation for signal and noise, time and frequency spatial double threshold-based denoising method is proposed, and it can differentiate the signal and noise effectively for both time domain and frequency domain; the basic theory of these three methods is introduced, also with the simulation experiment; the good experiment results proved the superiority of the time and frequency spatial double threshold-based denoising method.Finally, the technical for order identification in MGDM system is presented. Analyzing the property of ICA order identification in MGDM system educes that the key problem for ICA order identification in MGDM system is how to measure the correlation degree of signals separated by algorithm and detected signals of the system; concepts of correlation coefficient and correlation coefficient matrix which are used for describing the correlation of variable are introduced. Aiming at solving the ICA problem, this thesis defines the correlation coefficient matrix to describe the correlation of the separation signals and the mixed signals, also puts forward the order identification method based on the correlation coefficient matrix to diminish the problem of order indetermination. The thesis goes into particulars of implementation steps, after which the simulation experiment is carried out. The results from experiments prove the efficiency of ICA order identification method based on the mutual coefficient suggested by this thesis.Applying ICA in MGDM system can realize the demultiplexing only utilize the received signals. It also has many advantages such as without estimating the channel matrix, having the lower system complexity, easier to implement. And studies on the solution to demultiplexing and the order identification of ICA improve the demultiplexing performance and the selective ability of the route. These studies have an important significance in promoting the practical course of MGDM technology.