The Research On Brain-Computer Interface Based On SSVEP And Correlation Algorithm

In recent years,Brain-Computer Interfaces(BCI)as the rapid development of human-computer interaction technology which characterized is not depending on the normal output form of human body such as muscle,peripheral nerve of brain,bone,and so on.Directly controlled by the brain signal external devices.The research on brain-computer interface technology is at the stage of vigorous development,involving many fields such as life,medical treatment,military affairs,amusement and so on and plays an extremely important role.For some existed visual stimulation frequency Brain-Computer Interface Based on SSVEP system,SSVEP extraction can be considered as a classification problem,which can classify the SSVEP ingredients from the EEG signal.Although it can directly classify the time-domain EEG without any feature extraction and improve the arithmetic speed,too many time-domain EEG feature dimensions may cause "curse of dimentionless",To solve the problem,the article researches on the application of Locally Linear Embedding(LLE)in nonlinear dimensionless reduction of high-dimensional data,meanwhile in allusion to the problem of EEG signal's feature such as weakness and randomness,use spatial filtering to enhance the Signal to Noise Ratio(SNR),thus raise the classification accuracy,use a novel multiple electrode combination method was developed for SSVEP based BCI to enhance the Signal to Noise Ratio(SNR),thus raise the classification accuracy.The main work of this paper are as follows:Set up a based Brain-Computer Interface based on SSVEP system of the experiment platform.A complete experiment was designed to extract the SSVEP signal induced by stimulus frequencies.Then recruit some subjects to extract their EEG data at different visual stimulation frequencies.And analysis method of SSVEP signal and the prepossessing and spatial filtering of SSVEP signal are introduced in detail from three aspects: time domain,frequency domain and airspace.Aiming at the problem of excessive dimension of EEG signal,a local linear embedding algorithm is used to reduce the time domain feature of mufti-channel.And then,analysis based on the application of Locally Linear Embedding in SSVEP compared with classical features extraction such as Power Spectral-Density Analysis and Canonical Correlation.The results showed that the LLE had advantages in processing non-stationary EEG,In the short length of time window,the classification accuracy and information transmission rate of LLE was higher than other methods,in the slight variation of volatility period,with the increase of the time windows,the effect of LLE method was more obvious than others.In order to improve the performance of brain-computer interface system,the problems of parameter optimization and channel selection for each subject must be solved.In the SSVEP brain-computer interface system,the electrodes O1,O2,and Oz are frequently used electrodes,but because of the differences between subjects,the same electrodes are not optimal for all subject's effect,especially for some brain damage to the user,the brain part of the functional area may lose the original function,in order to allow brain-machine interface system for all users convenience,so the optimal electrode selection research is very necessary.In order to solve the above problems,this paper studies the optimal electrode selection for feature selection in brain-computer interface,and further improves the classification accuracy of EEG signals.Particle swarm optimization(PSO)algorithm was used to detect the optimal reference electrode,and the accuracy of the method was compared with that of single electrode,mufti-electrode,bipolar fusion and common average reference method in different electrode channels.Experimental results show that particle swarm optimization algorithm is superior to other methods in detecting the optimal electrode.