Research On Robotic Manipulator System Based On Motion Imagination EEG Signals

Brain-computer interface is one of the applications of combination of neuroscience and machine learning,which can m eet the control requirements of som e special environments,especially under the condition that the traditional contro l mode cannot meet the real needs,w here the brain-computer interface h as a unique advantage and extensive application prospects.This pa per mainly discusses the brain-com puter interface based on EEG signal.The em phasis is on extracting the patterns associated with brain motion imagination in EEG signals and identifying,digging and understanding the pattern characteristics of the brain under dif ferent active states.Finally,Platform is combined.In this paper,the basic theoretical framework of brain-computer interface based on EEG signal is introduced,in cluding the m echanism of EEG signal generation,the acquisition process of EEG signal,the unique ERS / EDS characteristic pattern and the preprocess of EEG signal.A set of filtering algorithms are design ed to extrac t the specific EEG signals of a specific freque ncy band from the intertwined EEG signals with a lar ge number of in terfering,and a denoising algorithm based on wavelet transform is designed to smooth the EEG signal.For the eigenvector feature extraction,a co-space pattern algorithm is introduced,based on which,an adaptive co-space pattern algorithm is proposed to solve the problem that the overall trend of EEG signal will be larger when the sample data is ex tracted.Because of the complex non-linear coup ling between the dimensions of multi-dimensional EEG,this paper introduces the principal component analysis algorithm to explore the principal component of EEG which will simplify the classif ication of multi-dimensional EEG sig nals.In the study of the classification of eigenvectors of EEG signals,classical linear discriminant analysis is used.At th e same time,consid ering the act ual situation of different EEG signals in different groups,an adaptive linear discrim inant algorithm is proposed to i mprove the original one.For the multi-dimensional and linear inseparable EEG signal classification problem,a nonlinear support vector machine with kernel function is introduced to make it linearly separable.For the multi-classification problem of EEG signal,the principal component analysis algorithm is used to d ecouple the EEG signal,and then the neural network classifier is designe d to c omplete the four classification tasks.Finally,the modeling and analysis of the two-link manipulator of the experimental platform are carried out,and it will b e used to implement the brain movement imagination task that identified by the Brain Computer Interface.