Blind Source Separation Application In EEG Signals Artifact Rejection

The core of the Brain-Computer Interface technology is to make the input EEG signals be converted into an output control signal. The EEG signals collected from the scalp of subjects is very weak and accompanied by a variety of artifacts,which brings great difficulties in extracting and analyzing the EEG data. Blind signal separation method is proposed under the case that the priori knowledge of the source signals is less known or unknown, the transmission channel characteristic is also unknown, EEG signals processing can use this model. For the artifact problem, this paper do a lot of research to find an automatic EEG artifact removal method based on the blind source separation way.Firstly, the research background and research status at home and abroad of the EEG signals processing are introduced, after then, this article give the basic knowledge of EEG and make a detailed description on the characteristics and classification of the EEG signals and the artifacts. Here we focus on the most serious ocular artifact and the 50Hz power line noise.Secondly,it introduces the basic idea of blind signal separation,whose mathematical model, constraint conditions and pretreatment process is used to solve the artifact separation.Also a deeply study of the classical blind signal processing algorithms (FastICA, JADE) are given. Here bring a new method named Stone's BSS into the brain signal analysis field,and there is no one used this method before, most of the previous works based on BSS have some inherent disadvantages. Stone's BSS method depends on signal temporal predictability measurement for separation processes. In addition, the proposed Stone's BSS here interpret how Stone's BSS deploys generalized Eigenvalue decomposition to obtain the un-mixing matrix based on the responses of two different linear scalar filters to the same set of signals.The paper also improves the Stone's BSS method to be a more mature way by the GA algorithm in the long and short filter part.The followed part in this paper,a comparison with some well-known BSS algorithms(JADE, FastICA) in isolating the simulation data in order to check the modified Stone's BSS effectiveness is given,which proves the modified Stone's BSS is better at the sub-gauss and gauss type signals.Finally, combine with the EEG and artifact characteristics and properties, make the modified Stone's BSS into practice to isolate the ocular artifacts, power line noise and correct the EEG data by a large number of experimental studies. And select the appropriate evaluation criteria to give the analytical judgment. The result shows that the three BSS algorithms make a good separation,in which the proposed modified Stone's BSS do the best.It is an efficient method to correct the EEG data, so people can apply it in medical applications as expected. Therefore, this paper opens a new direction in brain signal analysis field for Stone's BSS applications. At the end, give a final conclusion and also makes a directions for further studies.