The Key Algorithm Research And Blind Source Separation In Non-cooperative Communication System

Blind source separation(BSS)and its key algorithm are the focus of research in signal processing,image recognition,biomedicine and other fields at present and even in the future.This is because that BSS can obtain the expected results under unknown priori conditions.The main task of non-cooperative communication system is to separate the independent source signals from the received unknown mixed signals.Because the non-cooperative communication system does not know the prior knowledge of the received signal,so signal detection,parameter estimation,demodulation and identification processing are all“blind”and signal processing is always in a passive position,which make the non-cooperative communication receiving system can only separate the useful signals from the"blind"perspective by looking for the differences among time domain,frequency domain,space domain and code domain in cooperative communication and monitoring and intercepting those wireless signals lacking prior knowledge and obtaining various useful information carried by them so as to analyze and process the received signals more pertinently.In the process of signal interception and detection,the information in time-frequency domain and space domain is often the key,which is also the first signal technical characteristics to be determined for the non-cooperative communication system.Time-frequency characteristic represents the time domain and frequency domain properties of the source signal and space characteristics mainly refers to the direction of angle of the source signal(elevation angle and azimuth angle).In array signal processing,the spatial characteristics(elevation angle and azimuth angle)of the incident source signal can be obtained by using the estimation of the direction of arrival(spatial spectrum estimation).With the increasing complexity of electromagnet-ic environment,the characteristics of overlap each other in time\frequency\space and modulation domain bring great challenges to non-cooperative communication receiving system and unprecedented difficulty to signal separation.Therefore,it has become an urgent practical problem for non-cooperative communication systems to complete signal blind source separation and reconstruction in time-frequency domain and space(angle)domain.Based on the sparse representation theory and compressed sensing theory,a series of analysis and exploration were finished for the spatial spectrum estimation of non-cooperative signals,blind source separation under underdetermined conditions,sparse signal reconstruction and other problems.The theoretical framework of blind source separation and signal reconstruction in time-frequency domain was established,and a relatively perfect research scheme of blind source separation and signal reconstruction was constructed.The thesis mainly revolves around the following contents:The two-dimensional DOA(Direction of Angle)sparse estimation algorithm based on L shaped array is proposed to solve the problem of typical spectrum estimation in two-dimensional space.Based on the spatial sparsity characteristic of the cross covariance matrix,the sparse coefficients of the cross covariance matrix are found in the over-complete basis which is used to obtain the elevation angle estimation of the signal.Then the eigendecomposition of the conjugate transpose matrix of the data cross covariance matrix is carried out,and the azimuth estimation is realized by using the rotation invariant factor processing(ESPRIT-like method).During the estimation process,the elevation angle and azimuth angle can be matched automatically without additional parameter matching.Simulation experiments show that the proposed algorithm can achieve excellent performance of two-dimensional DOA estimation under the condition of low SNR and the DOA estimation error decreases with the increase of SNR as well as the number of snapshot gradually.The estimation algorithm shows good convergence performance.Compared with other algorithms mentioned in this paper,the DOA estimation accuracy is higher obviously,and the DOA estimation error is more close to the CRB value.The algorithm shows good estimation performance,Thus,the space angle information of the signal can be separated effectively.In order to solve the problem of underdetermined blind source separation(BSS)in non-cooperative communication,a new mixing matrix estimation algorithm is proposed based on the two-step method of BSS.On the basis of fully mining the angle information between the same array element signal and the energy information between different array element signals,a time-frequency domain single source point(SSP)detection strategy based on the double constraints of angle information and energy information is proposed,so as to realize the effective detection of SSP,and then FCM method is used to complete the mixing matrix estimation.By analysis and correlation simulation,the NMSE index of the estimation accuracy of the mixing matrix decreases with the increase of SNR,on the other hand,compared with the single-constrained SSP detection algorithm,the double-constrained SSP detection strategy can obtain a higher estimation accuracy in the estimation of the mixing matrix and can realize the accurate estimation of the mixing matrix.Secondly,to solve the signal reconstruction problem,a new underdetermined blind source reconstruction algorithm(SVMMUSR)with singular value membership matching is proposed.When the sparsity of the signal in the time-frequency domain change dynamically,by constructing the data expansion matrix and carrying out singular value decomposition and then detecting the membership of each data point to the detected subspace,the optimal k-dimensional subspace matching each data point under the condition of dynamic k-sparse is obtained and the subspace projection method is used to realize the accurate reconstruction of the source signal.The simulation results show that,compared with the conventional OMP,SL₀and TIFROM algorithms,the SIR index value of SVMMUSR algorithm will gradually increase with the increase of SNR,and the reconstruction performance of the algorithm is better than that of OMP,SL₀and TIFROM algorithms under the condition of high SNR.Meanwhile,with the gradual increase of SNR,the correlation coefficient of SVMMUSR reconstruction algorithm is approximately closed to 1 which indicating good signal reconstruction performance.Moreover,the performance of SVMMUSR algorithm is better than that of OMP,SL₀and TIFROM algorithm,which indicates that SVMMUSR algorithm has a higher signal reconstruction performance and can obtain a higher blind separation accuracy.Based on the research of underdetermined blind source separation for non-cooperative communication,the sparse signal reconstruction is further explored and discussed under practical conditions.In order to solve the problems of costing too long time and low precision in the sparse signal reconstruction algorithm when the actual sparsity is unknown,a matching pursuit reconstruction algorithm based on bidirectional sparse adaptive adjustment and weakly selected atoms matching pursuit(BSA-WSAMP)is proposed.In this algorithm,atom weak selection optimization strategy is adopted to optimize the update of support set,and“zoom”bidirectional variable step size method is adopted to realize the adaptive adjustment of sparsity,which can reduce the number of iterations effectively.According to the theoretical analysis,BSA-WSAMP algorithm does not improve the computational amount of signal reconstruction significantly but the reconstruction conditions allowed by the algorithm itself are more relaxed,which is more suitable for practical application.The simulation results show that,compared with other reconstruction algorithms Co Sa MP,SAMP,ASSAMP and SWOMP,BSA-WSAMP algorithm has higher reconst-ruction probability and faster algorithm convergence speed when the sparsity is large,and can achieve the specified reconstruction error when the number of iterations is small.At the same time,the algorithm has good adaptive characteristics for sparsity and has high sparse signal reconstruction quality while maintaining low reconstruction complexity and less reconstruction time.