Research On Key Technologies Of Group Brain-Computer Interface System

Human society determines that individuals cannot walk alone without groups,which is the embodiment of human sociality,and the result of individual development must be closer to the group.As a new approach for people to interact with the external environment,the study of Brain Computer Interface(BCI)has become matured.Common BCI interfaces includes: Motor Imagery(MI),motion visual evoked potentials(mVEP)and Steady-State Visual Evoked Potential(SSVEP);however,most of the current brain-computer interfaces are implemented based on single system and focus on the interaction between a single subject and the machine,which violates the essential attribute of people’s participation in group competition and cooperation.Among the signals used in the brain-computer interface,MI requires no stimulation but has a low signal-to-noise ratio,and the characteristics of mVEP and SSVEP are obvious but it is easy to be visually exhausted.A group brain-computer interface system integrating MI,mVEP and SSVEP was proposed,which was based on the advantages of three single-mode systems.According to the characteristics of the subjects,it can provide different signal forms of brain-computer interface,which allows the subjects realizing the coordination and competitive control of multiple users on external devices at the same time,stimulating the participation and enthusiasm of users,and finally it help realizing the brain-computer interface in the group sense.The research contents of this paper gives as follow.The key to the realization of the group system is based on the feature extraction and identification of a single brain-computer interface system,so it is necessary to develop a stable feature extraction algorithm for a single system.For the above three single-mode systems,SSVEP has high signal-to-noise ratio,which means its own stable and reliable signal recognition,while MI and mVEP have low signal-to-noise ratio,and the signal is easily interfered by noise.It is necessary to develop an effective identification method for MI and mVEP.Based on the good robustness of the fourth-order cumulant to gaussian noise,this paper developed the common space model based on the fourth-order cumulant for the extraction of MI and mVEP features.In this method,we first obtained the high-order cumulants of the relevant multi-dimensional eeg signals,and then realized the feature extraction of the Common Spatial Pattern(CSP)based on the high-order cumulants to serve as the features of the corresponding brain-computer interface.The method was applied to the recognition feature extraction of MI and mVEP.The results showed that the improved method could effectively reduce the influence of Gaussian noise on brain-computer interface feature extraction and significantly improve the performance of MI and mVEP brain-computer interface.Based on the study of feature extraction and recognition of single BCI system,we implemented a group online BCI system.This system was mainly based on the client server design pattern.with Cbuilder acting as the development tool.We using the existing low-level framework and Directdraw technology to achieve the front-end interface.The background mainly uses C++ as the programming language and uses multithreading and its interaction,memory sharing and other methods to achieve.Then use the socket communication protocol to implement client-side server communication,and realize the control of the corresponding game by the server through the message mechanism and the event interface.In this system,the client implemented the integration of MI,SSVEP and mVEP three modal brain-computer interfaces.Users can select the corresponding brain-computer interface according to their own characteristics We finally realize a two-person collaborate control of the game on the server.The online game control experiments of the group brain-computer interface system were carried out with 15 groups of subjects.The results showed that the subjects could perform the game confrontation operation more reliably under the selected brain-computer interface paradigm,and the system achieved the expected effect.