Research On Single-channel Blind Source Separation Method Based On Space Group Target Fretting Feature

In recent years,Blind Sources Separation(BSS)has been widely used in speech signal recognition,mechanical failure detection,brain wave pathology detection and radar signal target recognition.By means of matrix transformation or according to the probability and statistical characteristics of the source signal,each source signal is separated.Blind Source Separation(MCBSS)and Single Channel Blind Source Separation(SCBSS)can be divided according to the number of observed signals.Multi-channel blind source separation is the mixing of the dimensional source signals by some method,and the dimensional mixed signals can be obtained,and a source signal can be recovered from the dimensional mixed signals.Single-channel blind source separation is a kind of ill-posed problem of multi-channel blind source separation.After mixing multiple source signals,only one observation signal is obtained,and multiple source signals are separated from this mixed signal.Nowadays,multi-channel blind source separation has been widely used in the field of array antenna receiving signals,while single-channel blind source separation is very difficult to solve without prior conditions because of too little information,but it has higher research value because it is more suitable for practical application scenarios.In this paper,based on the group object composed of middle-section trajectory micro-warhead,multi-channel and single-channel blind signal separation technique and a signal de-noising technique are used as the line,and the echo separation method of the micro-group target is deeply studied in detail.It is mainly divided into the following aspects:In the first part,the research background and significance of blind source separation are introduced,and the research status of blind source separation is expounded.Combined with the research background of middle trajectory group target,the single channel radar receiving system is more practical,and the fretting form of warhead,as well as the mathematical model and separation evaluation standard of single channel separation are briefly introduced.In the second part,two different amplifiers are used to separate signals in the case of multiple channels.Firstly,the real part and the imaginary part of the complex echo signal are separated by the method of independent component analysis based on the negative entropy maximization of non-gaussian signal.Then,the fourth order cumulant matrix combined with diagonalization method is studied to separate the signals.By establishing the fourth order cumulant matrix and eigenvalue decomposition,the fourth order cumulant matrix is diagonalized to obtain the separation matrix and further separate the source signals.In the third part,the method of separating the echoes of different warheads and different fretting forms under the condition that the observed signals are single channel is studied.Firstly,according to the different fretting period characteristics of the warhead,the single channel observation signals are arranged into matrices,and the fretting period of each target is estimated by using the singular value ratio spectrum method.However,due to the serious time consuming of the algorithm caused by multiple singular values decomposition(SVD),the peak value satisfying the condition is extracted by the power spectrum method,and the reciprocal of the corresponding frequency is the target fretting period.Under the condition of different period of source signal,according to the difference of signal period,the source signal is reconstructed by using the maximum singular value obtained by SVD decomposition and the mixed signal is separated by piecewise superposition method,and the separation results are compared by simulation.Under the permise of accurate estimation of the period,the singular value method can only get the approximate waveform of the signal,and the separation error is large,and the signal accuracy of the segmented superposition method is higher.In the forth part,a series of optimization problems for denoising mixed signals with Gaussian white noise are studied,and mixed signal after denoising is separated by the second part method.Firstly,according to the distribution characteristics of white noise,on the basis of empirical mode decomposition(EMD),different methods for finding the criterion of noise and signal boundary point are studied,and the denoising method similar to wavelet soft threshold denoising is used for noisy components.Due to the problem of modal aliasing in EMD method,the effect of signal separation after denoising is not ideal.The complete empirical mode decomposition(CEEMDAN)method can effectively improve the modal aliasing problem,improve the accuracy of denoising,and thus reduce the error of separated signal.The above principle is verified by simulation.