Based On Blind Source Separation Research Of EEG Artifacts Remove

Blind signal separation(BSS)is a powerful signal processing method developed rapidly in the late twentieth century.As a product of artificial neural network,statistical signal processing,information theory and computer,Blind Source Separation has become the research and development in many fields.Especially in biomedicine,image processing,remote sensing,radar and communication systems,earth sciences,data mining and other aspects have the outstanding contributions.EEG is an important physiological signal produced by the human brain.EEG signal is very weak and has a non-stationary,in which EOG artifacts is the main source of measured EEG signal interference.As the EEG signal and EOG artifacts are independent of the signal source,so blind source separation method to remove EOG artifacts can be a good solution to such problems.In this paper,the characteristics of EEG and its acquisition method are discussed.The blind source separation algorithm of EEG signal artifacts in positive definite case is simulated.Underdetermined blind separation algorithm is studied.Based on k-means clustering algorithm improves the accuracy of the hybrid matrix.The specific work is done as follows.1.According to the characteristics of EEG and the requirement of separation algorithm,the EEG signal has its own characteristics.Before EEG signal separation,the EEG signal should be analyzed,which greatly contributes to the understanding of the algorithm characteristics and the reduction of the computational complexity of the algorithm.2.For the blind source separation of EEG signal in positive definite case,we mainly study fixed-point algorithm and natural gradient algorithm.The calculation model based on fourth-order cumulant and negative entropy is analyzed and the artifacts are separated by simulation.The natural gradient algorithm is found to make the EEG signal more pure,but the number of iterations is relatively It is found that the number of iterations is reduced and the separation efficiency is improved.This method achieves the purpose of removing the pure electroencephalogram signal by removing the artifacts.3.According to the characteristics of EEG signal and the sparsity analysis,the two-step method in sparse component analysis is used to solve the problem of EEG signal acquisition channel number less than the original signal number Happening.4.For the two-step method,three algorithms,k-means algorithm,improved potential function method and improved k-means clustering algorithm,are used to estimate the hybridmatrix.The improved k-means clustering algorithm improves the accuracy of the estimation of the hybrid matrix.The shortest path method is used to estimate the source signal.Simulation results show that the improved algorithm improves the precision of EEG signal separation.