Design And Implementation Of The Brain-Computer Interface Based On Brain Network Techniques

Brain-computer interface(BCI)is a wholly new system which does not rely on the peripheral nervous and muscle tissue and is completely autonomous to realize the direct control of the human brain to the external environment.BCI is an interdisciplinary technology which involves many subjects and has a broad prospect in the fields of medical rehabilitation,sports aid,human-computer interaction and so on.With its advantages of simplicity and high time resolution,EEG(electroencephalogram)is one of the most widely used technologies in BCI system.However,the study found that about 15% to 30% of users exist “BCI blind” problem.In such cases it is difficult for the users to induce a strong(Event-Related De-synchronization)ERD /(Event-Related Synchronization)ERS signal,which means that we cannot measure the relevant rhythm signal.In addition,there are differences in the characteristic signals of different subjects.All the problems mentioned above led to the fact that EEG-based BCI requires rigorous screening of subjects and extensive pre-training.However,the EEG time series reflect the synergistic relationship between different brain areas.Based on this,the EEG time series can be converted into a network whose measures are associated with consciousness.The idea of converting EEG time series into a network provides a new way to solve the problems of BCI.With the aim to design and implement the asynchronous BCI system,the paper takes the brain network technology as the breakthrough point and focuses on solving the signal processing problem in the BCI implementation process.The main work of this paper is as follows:(1)In order to solve the problem of how to establish the brain network model,first is positioning and activating the brain region to determine the 18 key channels such as FC1 as the brain network nodes.Then adopting the coherence coefficient(COH)and phase locked phase(PLV)from the amplitude and phase angle respectively to quantize the relationship between nodes and build weighted and binary networks.(2)Around the core issue of differentiating the working-state and resting-state in asynchronous BCI,the EMG signal is used to determine the motion start time and the time period of 500 ms to 1200 ms before the motion is extracted as the bereitschaftspotential state,which replaces the EEG in the exercise execution to differentiate the working-state and resting-state,reducing brain-network-technology based BCI response delay.(3)Aiming at the key problems in asynchronous BCI implementation,the experimental paradigm for brain network technology is designed,and the problem of signal preprocessing is solved,the characteristics of brain network are extracted,and the characteristics are identified by linear Fisher and nonlinear SVM classifier.The correct rate of classification based on Fisher and SVM,the COH weighted network is 61.70% and 67.70% respectively,and the correct rate of binary network classification is 69.48% and 69.75% respectively.The correct rate of classification based on PLV weighted network is 64.78 % and 69.47% respectively,and the correct rate of binary network classification is 72.28% and 76.72% respectively.(4)Based on the algorithm mentioned above,we choose the clustering coefficient of the PLV binary brain network as characteristic and SVM as the main classifier to realize the online preprocessing,feature extraction and classification recognition algorithm.The experimental results of the asynchronous BCI system show that the system can determine the working-state and resting-state and the online recognition rate reaches 59.84%.The results show that the asynchronous BCI system based on brain network technology is feasible and can be used as a new way.