Research On The Method Of Mixed Signal Separation Based On Variational Mode Decomposition

Single channel blind separation is a kind of blind source separation(BSS).It is a important issue in the field of signal processing.The blind source separation represents the separation of the source signal from the observed signal in cases where the prior information is minimal and original signal,mixing type,channel information,etc.are unknown.At present,blind source separation algorithm is usually used to solve linear determined problems.While the under-determined problem are still in development.Although the problem of single channel blind separation is a case of under-determined blind separation problem,the NMF algorithm has a bad result when it is used to sovle the problem of single channel blind separation.The research of this paper focuses on single-channel blind separation.The core contents have the following two aspects:(1)The number of intrinsic mode functions decomposed by the empirical mode decomposition(EMD)can not be determined.And the mode mixing is easy to occur.Therefore,the single-channel blind source separation algorithm based on empirical mode decomposition is unstable.To solve the above problems,we choose the variational modal decomposition(VMD)algorithm instead of the empirical modal decomposition algorithm.It converts the single channel into the virtual multi-channel,so that the ill-posed problem is transformed into the normal problem.Finally,the fast independent component analysis(FastICA)algorithm is used to separate the source signal.The simulation results show that the algorithm can effectively separate the source signals.The separation performance is better than the separated signals based on empirical mode decomposition.(2)The non-negative matrix factorization(NMF)algorithm is be well to solve the underdetermined problem.Then,we can try to apply it to single-channel blind source separation.The results show that the effect of this experiment is not stable.So,a improved method is proposed to solve this problem.The algorithm improves the correlation between the source signal and the separated signal.In this paper,we mainly study the above two aspects,and use a large number of mathematical models and simulation experiments to show the feasibility and effectiveness of these algorithm in the waveform of signal separation.