Based On Blind Source Separation Research Of The Ocular Artifacts Remove

EEG which is always spontaneous electrical activity of the central nervoussystem, is generated by the brain activity， so it is an important human physiologicalsignal. The randomness and a very strong non-stationary of EEG is very weak, andhas more complex background noise. EOG artifacts with high frequency andamplitude which has become the main interference sources of the actual EEG, isseveral times of the EEG. Therefore, the removal of ocular artifact interference andthen extract pure EEG, have important theoretical and practical value for practicalapplication of the brain computer interface.EEG and EOG artifacts are derived from independent sources, so we chooseblind source separation method for removing ocular artifacts interference as animportant entry point. Aiming at the current problems for blind source separationmethod in removing EOG artifacts, the problems including the need for manualidentification, it could easily lead to a potential electrical components loss of EEGand other ingredients. We take the blind source separation as the foundation, and thenput forward new methods for the removal of EOG artifacts automatically with highaccuracy. The new methods have great significance for the developmentand application of real-time Brain Computer Interface system. The main researchresults include:(1)First, we explain the basic processing mechanism to remove eye blind sourceseparation based on blind source separation from research methods, calculation model,pretreatment, evaluation indexes four aspects. The basic processing mechanismprovides theoretical foundation and algorithm supports for the automatic removal ofEOG artifacts with high precision.(2) The new method use the independent component analysis and second-orderblind identification as the foundation, according to the problem which is based onblind source separation method and requires human identification in EOG artifactsremoval, then we propose a new method based on the combination of second-orderblind identification and nonlinear parameters. Firstly, the method separates EOG artifacts interference from the original EEG signal through the second-order blindidentification way, then the combination of Nepal entropy and sample entropy aretaken as the discriminate factor, and a universal artifact identification method is usedto identify EOG artifacts, which realizes the EOG artifacts automatic identificationand removal; finally, we reconstruct the pure EEG signal. On this basis, we design theexperiments of automatic EOG artifact removal, and then estimate EOG artifactsthrough the correlation coefficient, signal-to-noise ratio and the running. The resultsshow that the method can remove the EOG artifacts automatically and effectively.(3)On the basis of the above research, aiming at the problem that the loss ofthe EEG power spectrum during the ocular artifacts removal, we make full use of thedifference between EEG and EOG artifacts spatial features, then a new EOG artifactsremoval method based on the combination of Second Order Blind Identificationand canonical correlation analysis is put forward in this paper.(3)Firstly, we get the decomposition signal through Second Order BlindIdentification which is based on the theory of blind source separation; at the sametime, in order to obtain high-precision ocular artifact removal, the eye referencechannel signals are combined with the decomposition of the signal to obtain acombined signal. Secondly, we extract the canonical correlation variable namelyartifact component through canonical correlation analysis method. Then, in order toreduce the loss of EEG spectrum components, the new method is used four differentways to process the extraction of the canonical correlation variables. Finally, themissing components of EEG signal are recovered through the signal reconstruction.On this basis, we use signal to noise ratio indicate EOG artifact removal, and theextent of the correlation coefficient expresses the missing of interested EEG single,then the running time is estimated using the real-time nature. The results show that,on the basic of EOG artifacts removal effectively and automatically, the new methodcan keep the useful brain power signal as much as possible.