Research On Multi-object Resolution Algorithm Based On Single Channel Blind Source Separation

Blind Source Separation(BSS) is a kind of Blind signal processing,which has developed rapidly in recent years.It is widely used in the fields of mechanical fault diagnosis,geological exploration and detection,as well as biomedical,voice communication,array structure and other signal processing fields.Because the single-channel blind source separation problem is common in the actual environment,the single-channel blind source separation algorithm has been widely concerned by domestic and foreign scholars in recent years.and the single-channel blind source separation algorithm has a strong practical value and an optimistic research prospect,so it is very necessary to study and explore it.This paper mainly studies multi-channel BSS algorithm and single-channel BSS algorithm.On the basis of single-channel BSS algorithm,it studies vector MUSIC algorithm for azimuth estimation.Besides,The algorithm is analyzed and verified by simulation and experimental data processing results.The paper firstly introduces the basic theory of blind source separation model,analyzes the two uncertainty factors and the hypothesis of signal model in BSS problem,and studies the pretreatment process of BSS algorithm.Then,it analyzes the relationship with independent component.Finaly,it introduces several separation performance evaluation index.The above content lays the theoretical foundation for the following algorithm research.The multi-channel BSS algorithm mainly studies the information maximum(Informax)algorithm and fast fixed point(Fast ICA)algorithm.This paper compares and analyzes the separation effect of the two algorithms by simulation and experimental data processing.It is found that under the same conditions,the Fast ICA algorithm is more stable than the Informax algorithm,with fast convergence speed and strong applicability.However,when the number of signal sources is large and the signal to noise ratio(SNR)is low,the iteration times of Fast ICA algorithm increase and the performance deteriorates.On the basis of the real signal fast-ica algorithm,the complex form of the algorithm is further studied,and the azimuth information of the signal source is derived through the estimation of the mixed matrix.Finally,it analyzes the accuracy of azimuth estimation under different SNR.The single-channel BSS algorithm mainly studies the single-channel blind sourceseparation algorithm based on wavelet packet decomposition(WPT-ICA),single-channel blind source separation algorithm based on empirical mode decomposition(EMD-ICA)and single-channel blind source separation algorithm based on variational mode decomposition(VMD-PCA).Based on the VMD-PCA algorithm,the vector MUSIC algorithm is used to estimate the azimuth,which is called VMD-PCA-MUSIC.Through wavelet packet decomposition and reconstruction of observed signals,WPT-ICA algorithm forms a new multi-channel signal with high energy reconstructed signals and observed signals,and then conducts blind separation with Fast ICA algorithm.This algorithm can effectively solve the problem of single-channel blind separation.The emd-ica algorithm has the phenomenon of mode aliasing in the separation of mixed signals,which requires that the frequencies of the source signals meet a specific relationship,or the algorithm fails to separate.Although the eemd-ica algorithm can solve the mode aliasing to some extent,it can't eliminate it completely,and it has many iterations and takes a long time.Compared with the above algorithms,VMD-PCA algorithm has the lowest sensitivity to noise and the best effect.Based on VMD-PCA algorithm,VMD-PCA-MUSIC algorithm which combines VMD-PCA algorithm with vector MUSIC algorithm is proposed in this paper.This algorithm effectively solves the problem that MUSIC algorithm cannot provide correct orientation information for mixed signals with different frequencies.Finally,the directional accuracy of VMD-PCA-MUSIC algorithm and vector MUSIC algorithm for mixed signals is analyzed and compared.Simulation results and experimental data show that the VMD-PCA-MUSIC algorithm has higher orientation accuracy.