Study And Implementation Of Hybrid Brain Computer Interface System Based On SSVEP And Motor Imagery

In recent years,researchers have proposed a direct communication channel between the brain and peripheral devices,in order to improve elderly people and paralyzed people to live independently.The new technique is brain computer interface(BCI),which has wide good application prospect in the field of neural engineering.The current BCI research is mainly based on a single type of BCI system,which can only identify simple tasks and can not meet the requirements of practical applications.The hybrid BCI system that integrates different types of EEG signals can solve this problem,Therefore,the hybrid BCI system is the focus of the present research.This thesis investigates a hybrid BCI system based on steady-state visual evoked potential(SSVEP)and motor imagery(MI),which increases the number of identification tasks and completes the online control of the Dobot manipulator writing experiment.The main research work of this paper is described as follows:(1)The design of experiment.This paper designs SSVEP experiment paradigm,motion imagination experiment paradigm and hybrid experiment paradigm respectively.In the MI experiment paradigm,a friendly prompt interface is designed.In the SSVEP experiment paradigm,the effects of stimulation frequency and electrode position on the SSVEP characteristics were studied under the factors of stimulus mode,stimulus target size and stimulation target layout.And the optimal results in the SSVEP experimental paradigm are used to design the hybrid BCI paradigm.(2)The EEG features are extracted from different types of EEG signals.In the MI paradigm,the original signal was preprocessed,and the signal was extracted by using the short time fourier transform method(STFT).The results showed that the effectiveness of the electrode channel and the mu rhythm had obvious ERD/ERS phenomenon.And the feature extraction of EEG signals is realized by the mu rhythm second moment energy method,which saves the processing time.In the SSVEP experiment paradigm,the low frequency interference part is removed by wavelet denoising method,and the signal is processed by the improved canonical correlation analysis method(CCA)according to its obvious frequency characteristics,and the optimal parameters and thresholds are analyzed.Compared with the traditional power spectral density analysis method(PSDA),the results show that the CCA method is more effective in extracting signal characteristics.(3)The classification of EEG signals.This thesis studies linear classifier and support vector machine(SVM),SVM based on RBF(radial basis function)as classifier,the cross validation algorithm is used to solve the selection of penalty factor and kernel function,which improves the performance of the classifier.And Experimental results show that different classifiers have good classification accuracy.(4)A novel hybrid BCI system was proposed to achieve multiple command controls ofthe Dobot robotic arm in three-dimensional space(3D).In the online hybrid BCI system,six subjects participated in this study and succeed to manipulate a Dobot robotic arm in 3D space to write some English letters.And the mean decoding accuracy of writing task was 89.73±2.77 %.In summary,the hybrid BCI system based on SSVEP and motor imagery enables online control of Dobot robotic arm.The result suggested that our proposed non-invasive hybrid BCI system was robust and stable.At the same time,it provides a new idea of the application of BCI technology in home environment.