| Brain-Computer Interface(BCI)technology as a new type of human-computer interaction technology that doces not rely on peripheral nerve and musclce,and which can provide a kind of information communication between brain and external devices for people with physical disabilities Who have normal nerve activities.In addition,BCI is a potential and promising technology in fields of military and medical rehabilitation,By the means of collecting the Electroencephalogram(EEG)signals generated by the different thinking activities of the brain,and making classification of EEG mental tasks possible through physiological phenomena of different EEG rhythms.In the mental task-based BCI system,we collected the EEG signals related to the left and right hand motor imagery and mathematical formula calculation of users by using the EEG acquisition equipment,then extracted and classified the feature vector of data,and combined different mental tasks with computer insitructions,so as to achieve the communication between the brain and external devices.In EEG-based mental tasks,feature extraction and recognition classification of EEG signals are the most critical procedure,if the user's thinking intention cannot be correctly interpreted.the external equipment cannot be properly controlled.Common Spatial Pattern(CSP)is a feature extraction method widely used in EEG signals,which can effectively extract corresponding features.Based on the traditional CSP algorithm,this thesis proposes an incremental updating algorithm for CSP filter based on training sample evaluation to solve the shortcomings that CSP online system needs to collect a large number of training samples and cannot judge the quality of data.According to the advantages of simple design and high recognition rate of CSP algorithm,the feature extraction and classification of five kinds of mental tasks were carried out.The experimental results showed that CSP algorithm can effectively classify ulti-class of mental tasks.This method has certain application prospects in BCI based on mental tasks.The main work of this thesis includes:(1)The physiological mechanism and main rhythm of EEG signals were studied.The feature extraction methods such as thr frequeney donmain,the time domain,thespatial domain filtering,and classification algorithms were used to give an overview of EEG signals.(2)An incremental updating algorithm for CSP filters based on EEG training sample evaluation was proposed.By screening and eliminating "low-quality" data for a single EEG sample of motor imagcry,and combining with the incremental CSP algorithm,the problem that it is difficult to obtain the optimal CSP filter by manually screening EEG samples and a small amount of training data in the experiment of CSP-BCI system was solved.Experimetal results show that this method improves the performance of BCI system.(3)The application of CSP algorithm in multi-task EEG signial classification was studied.A five-class EEG mental tasks experimental paradigm was designed.Based on the two-class CSP algorithm,pre-processing,channel optimization and feature extraction were performed on mental tasks.Through the "one-to-one" CSP method-multiple CSP filters were constructed to map the common spatial mode of EEG signals,and the projection signals of each kind of mental tasks were obtained,the variance of the projection signal energy was obtained as the feature vector,and finally the recognition and classification were carried out.The results showed that the CSP algorithm has a good effect in the classification of multi-class of mental tasks.(4)A common spatial pattern offline application system was designed under the MATLAB platform.The system mainly implements two functions of EEG sample quality screening and individual identification.Seven subjects participated in the individual identification experiment.The experimental results showed that the system can accurately identify the identity of the subject by the EEG data when they were at rest with eyes closed. |
PDF Full Text Request