Research And Application Of Brain Network And Brain Computer Interface

With the rise and development of brain science,complex brain networks are increasingly used in the study of the brain,and brain-computer interface,a new type of technology that enables the brain to communicate directly with the outside world,has also emerged.With brain-computer interface,it is possible to directly achieve "brain control" of peripheral devices without the assistance of limbs.Steady-state visual evoked potential(SSVEP)is widely used in this field,which is a kind of EEG signal aroused by visual stimulation at a specific frequency.In this thesis,a number of EEG acquisition experiments based on SSVEP were designed,and the EEG data of several subjects was collected with the help of related equipment.The research contents are as follows:(1)Based on the relevant knowledge of complex network theory and graph theory,this thesis uses EEG data to construct a functional brain network based on coherence and a brain network based on mode transfer,and uses a series of network topology properties and significance testing methods.After analysis and comparison,the results showed that the SSVEP response of the occipital and apical regions of the brain was more significant,and the SSVEP response amplitude was different among different subjects.Moreover,the larger the SSVEP response amplitude was,the larger the clustering coefficient,global efficiency and local efficiency of the coherence based functional brain network was,while the smaller the characteristic path length was.In addition,the average small-worldness of all subjects in the fatigue state increased,indicating that entering the fatigue state will make the brain's mode transition more frequent,and the connection of the local network in the occipital region of the brain will be enhanced.(2)This thesis analyzes a variety of feature extraction and classification algorithms,and after analysis and improvement,a hybrid analysis model based on canonical correlation analysis,filter bank canonical correlation analysis and task-related component analysis is proposed.It and other basic algorithms are used to process the collected offline EEG data.After comparative analysis,the hybrid model can achieve higher average recognition accuracy and information transmission rate in a relatively short period of time,and its performance is better than other single methods.(3)This thesis designs and implements an online brain-controlled game system.These include writing a visual stimulation module for SSVEP control and experimentation,a data acquisition module based on a portable EEG amplifier,a realtime signal processing module based on a hybrid model,and a control and application module composed of a self-developed 3D maze game.It has been verified by online experiments that the average recognition accuracy of the online brain-controlled game system has reached 81.46%,and the average information transmission rate has reached20.29 bit/min,which can basically meet the requirements of accuracy and real-time performance.