The Research Of Single-channel Blind Source Separation Algorithm

Blind source separation(BSS)problem is the process of separating the source signals from the observed signals in case of where both the transmission channel and the source signals are unknown.BSS algorithms have been applied in many fields,such as biomedical signal processing,array signal processing,mobile communication and so on.In recent years,single-channel blind source separation(SCBSS)problem has become a hot research topic in signal processing field.SCBSS is an extreme case of underdetermined blind source separation problem.It is difficult to separate multi-source signals by using the characteristic information of single-channel observation signals.However,it is a typical problem of some practical system.Thus research of SCBSS algorithms is of important theoretical significance and application value.This paper mainly studies on BSS algorithm and SCBSS algorithm.Firstly,the improvement of FastICA algorithm is studied.Aiming at the problem that when the number of source signals is large,the number of iterations of the original FastICA algorithm is increase and the separation performance is worse,the Pm-FastICA algorithm is proposed.The nonlinear function of the algorithm is approximated,and the rational function can be obtained which can reduce the iteration times of FastICA algorithm,thus the convergence speed and separation performance are improved.The simulation results show that the performance of Pm-FastICA algorithm is better than that of FastICA algorithm,and the performance advantage of Pm-FastICA algorithm will be more obvious as the number of source signals increases.At the same time,a FastICA algorithm using a rational polynomial nonlinear function(N-FastICA)is proposed.The simulation results show that the performance of N-FastICA algorithm is better than Pm-FastICA algorithm and FastICA algorithm,and the performance advantage of N-FastICA algorithm will be more obvious as the number of source signals increases.Secondly,the SCBSS algorithm based on wavelet packet decomposition(WPT-ICA)is studied.In accordance with that the performance of SCBSS algorithm based on wavelet transform has yet to be improved,because the wavelet transform can not analyze the detail information of a signal well,a SCBSS algorithm based on wavelet packet decomposition is presented.The wavelet packet decomposition of the observed signal is carried out,and the coefficients with higher energy percentage are reconstructed.The reconstructed signals and the observed signal are composed of multiple signals,and the source signal is separated by N-FastICA algorithm.The simulation results show that the SCBSS algorithm based on wavelet packet decomposition is superior to SCBSSalgorithm based on wavelet decomposition.Thirdly,the SCBSS algorithm based on Empirical Mode Decomposition(EMD)is studied.Since EMD-based SCBSS algorithm exists modal aliasing phenomenon,its separation performance is deteriorated.Thus a single-channel blind source separation algorithm based on EMD,Principal Component Analysis(PCA)and Independent Component Analysis(ICA)(called EP-ICA algorithm)is proposed.The intrinsic mode function(IMF)components are decomposed by empirical mode decomposition.For the IMF components with modal aliasing,the multi-channel signals are constructed by using its periodicity,and the aliasing is eliminated by ICA and the fake components are removed by using PCA and the cross-correlation.And then the new multi-channel signals are got by the remaining signals and observed signal and the source signals are separated by using the N-Fast ICA algorithm.The simulation results show that the EP-ICA algorFinally,the SCBSS algorithm based on Variational Mode Decomposition(VMD)is studied.VMD is introduced into the SCBSS algorithm,at the same time,the feedback mechanism is applied to the VMD method,and the VMD-SCBSS algorithm and an algorithm of SCBSS based on feedback VMD(VMDF-SCBSS)are proposed.The simulation results show,that the VMD-SCBSS algorithm and VMDF-SCBSS algorithm are superior to EP-ICA algorithm,the VMDF-SCBSS algorithm has the same performance as the VMD-SCBSS algorithm.The VMDF-SCBSS algorithm can effectively determine the source signal number,and its computational complexity is lower than that of the VMD-SCBSS algorithm,and need not to predict the spectral interval.