Research On Modulation Identification Technology Of Wireless Communication System

The technology of signal modulation system recognition is a key tech-nology in non-cooperative communication receivers,which is very significant in adaptive modulation,spectrum management,spectrum auditing,electronic countermeasure,and military reconnaissance.In the current non-cooperative communication signal modulation system identification,there are many prob-lems.For instance,in the complex channel environment,the signal recognition performance is very low.In addition,obtaining partial signal samples is very difficult,which cannot meet the practical application requirements.Aiming at those problems,this thesis conducts the research on the key technologies of modulation system identification in non-cooperative communication systems,which mainly involves the identification of mobile communication signals in non-Gaussian information channels,the recognition of MIMO signal in cor-related decline channels,the identification of digital modulation signal with fewer signal samples,and the recognition of the intra-pulse signal of radar in short wave channel.The main findings and contributions of this thesis are as follows:· There are many problems in the characteristic extraction of traditional mo-bile communication signals.For instance,the characteristic differenti-ation is not obvious in low Signal-to-Noise Ratio(SNR)and multipath time-delay channels.Moreover,a large amount of calculation is needed.Aiming those problems,this thesis puts forward a signal characteristic extraction method on the basis of the instantaneous feature extraction,which effectively overcomes the noise interference.The simulation re-sults indicate that this method not only own obvious feature differentia-tion,but also own low computational complexity.In the traditional mul-tipath time-varying channel environment,there are many problems,such as poor recognition performance,low recognition rate,and high compu-tational complexity.Aiming at those problems,this thesis puts forward a machine learning algorithm on the basis of the ELM extreme learning machine.The algorithm generalizes the signal recognition problem into multivariate characteristic classification problem.Moreover,this algo-rithm does not require any prior infonnation.In addition,it owns better recognition performance and lower computational complexity in the en-vironment of non-Gaussian time-varying multipath decline channels.The simulation results display that the method possesses good recognition per-formance in multipath channels with low SNR.Aiming at the problem of limited recognition performance of ELM extreme learning machine in ab-normally bad channel(such as low SNR,increasing multipath number,and increasing time-delay),this thesis proposes the SAE-ELM adaptive machine learning algorithm.This algorithm optimizes the network param-eters in the machine learning process through differential mutation evo-lution theory,which effectively overcomes the influence of the channel on the signal in the bad environment.The simulation results display that this algorithm owns good recognition performance in complex channel environment under the premise of increasing the complexity of a certain amount of the algorithm.· Aiming at the problem that the identification performance of the MIMO communication signals in low SNR and multi-path channel,this thesis puts forward a digital signal characteristic extraction method on the basis of the cyclic stationary characteristics under the relevant MIMO channel model.Theoretically,this method can eliminate the influence of noise and multipath decline channel.Moreover,through simulation,its relia-bility of the feature extraction method is verified;aiming at the current problem of long operation time and low recognition rate of existing clas-sifiers,this chapter also introduces the ELM machine learning algorithm as well as the SAE-ELM machine learning algorithm.The results show that ELM machine learning algorithm,affected by low SNR and multipath decline channel,can achieve high accuracy of signal recognition;aiming at the problem of the limited recognition accuracy of ELM algorithm in bad channel environment,this thesis proposes a MIMO signal modulation recognition method on the basis of the channel equalization.In the case of low channel estimation error,the influence of multipath channel and noise can be completely eliminated.The results reveal that the introduc-tion of equalization algorithm greatly improves the recognition accuracy of MIMO signal modulation system,and provides a new research mental-ity for signal recognition in bad environment.· Aiming at the problem that a large number of data samples are needed to train classifiers in the traditional signal recognition process,this thesis puts forward a SS-ELM semi-supervised learning machine algorithm un-der the condition of fewer data samples.This algorithm effectively over-comes the problem of classifier algorithm in the case of fewer data sam-ples.Theoretically,it analyzes the feasibility of signal recognition and classification under less signal data samples.Moreover,its effectiveness in signal recognition is verified by experimental simulation.The simula-tion results reveal that the method owns better recognition effect in low SNR.Meanwhile,this thesis conducts the research on this algorithm in dif-ferent labeled samples and puts forward a method of stair equal interval of sample number.The results show that there is a correlation between the number of training data samples and the performance of semi supervised machine learning algorithm.With the increase of training data samples,the recognition performance of the classification device is improved ac-cordingly.· Aiming at the problem of poor recognition performance of traditional in-trapulse radar signals in short-wave non-Gaussian multipath time-delay channels,this thesis puts forward a method of radar signal modulation recognition on the basis of wavelet transform.Through extracting the pa-rameter feature of wavelet transform domain as the recognition feature,this method effectively achieve the radar signal modulation system recog-nition under the non-Gaussian multipath time-delay channel.In the selec-tion of classifiers,ELM and SAE-ELM machine learning algorithms are selected as feature classification methods.The simulation results display that this recognition method not only owns better performance than the traditional one,but also owns a high recognition rate and good robustness in shortwave channel.Finally,the content of this thesis is summarized.Moreover,the development direction and future work of wireless modu-lation system recognition are prospected.