Research And Application Of Signal Parameter Estimation And Modulation Classification In Non-cooperative Communication

The rapid development of communication technology in the information age,the increasing diversity of application scenarios,the richness of communication signal types,and the complexity of communication environments have all put forward higher requirements on communication technology.In the non-cooperative communication,which lacks priori knowledge,the challenges are even greater,and the study of non-cooperative communication systems has important theoretical and application values.This thesis mainly focuses on the research work on the estimation of signal parameters and the classification of signal modulation in non-cooperative communication.In the estimation of signal parameters,this thesis firstly designs a SNR estimation algorithm based on autoencoder for Gaussian noise,which does not require different spectral estimation window lengths according to different symbol rates compared to the algorithm based on power spectral distribution function inflection detection,and the algorithm has better accuracy and stability,and it is applicable to the four major types of digital modulation signals studied in this thesis,and it is robust to symbol rates and roll-off factors.For the unavoidable frequency offset in the non-cooperative communication system,this thesis proposes an estimation algorithm based on artificial bee colony,and a new feature parameter-intensity of probability distribution of amplitude and phase ρAP is proposed to measure the effect of the frequency offset estimation,and the larger ρAP is,the better the frequency offset removal effect.Considering the conflict between accuracy and computational effort in the traditional iterati ve solution of the optimal problem,the artificial bee colony algorithm is used to solve the maximum value and the corresponding optimal frequency offset.Compared with the clustering algorithm based on grid,the proposed algorithm does not need to set the threshold and has higher estimation accuracy at low SNR.For the three kinds of digital modulation signals studied in this thesis except MFSK signals,the frequency offset estimation algorithm performes well in different roll-off factors and symbol rates.In the classification of signal modulation,this thesis realizes digital modulation signal classification by analyzing signal transient features and spectral features.Firstly,a classifier based on decision tree is designed,which adopts the classification scheme of constant envelope and non-constant envelope signal interclass classification first,and then intraclass classification.For MFSK signals,a classification algorithm is proposed to extract the main instantaneous frequency distribution using wavelet threshold denoising and kernel density estimation,through the way with template matching,which is applicable to the classification of MFSK signals with smaller modulation indices in real systems compared to the algorithm that counts the number of spectral peaks of the signal spectrum to distinguish the signal types.For 16QAM and 64QAM classification,a new feature parameter is constructed-the ratio of the variance to the mean squared of the normalized instantaneous magnitude probability density distribution.It is experimentally verified that the modulation classification algorithm based on decision tree has an accuracy of more than 90%for 16 kinds of digital modulation signals when the SNR is higher than 11dB.However,the effect is unsatisfactory at low SNR.To solve this problem,based on the feature parameters selected from the decision tree,combined with the MFSK signals classification algorithm proposed in this thesis,the modulation classification scheme based on random forest is designed,and the classification accuracy can exceed 90%when the SNR is higher than 8dB,and the classification effect is significantly improved at low SNR.Finally,the system implementation of the designed algorithm scheme is carried out.The main work includes porting algorithm and designing software,and the test platform is built and the parameter estimation and modulation classification modules are tested for the requirements of the national major instrument research and development project-Development and Application of Illegal Electromagnetic Signal Monitoring and Classification Technology.From the test results,it can be seen that the algorithm and system implementation proposed can achieve good results in real electromagnetic environment.