Brain-computer Interface Research Based On High Frequency Steady-state Visual Evoked Potential

Brain-computer interface(BCI)can be defined as a new type of communication control system built between the brain and electronic equipments represented by computers.BCI technology involves neuroscience,computer science,control theory,and many other technologies.It has extremely high application value and broad application prospects in many fields such as life,entertainment,medical,aviation,military,and rehabilitation engineering.Due to the different ways of obtaining EEG signals,BCI can be divided into invasive BCI and non-invasive BCI.As a non-invasive EEG signal,steady-state visual evoked potential(SSVEP)is a rhythmic response recorded in subjects’ brain scalps when exposed to external visual stimuli that flicker at a constant frequency.BCI based on SSVEP has the advantages of high information transmission rate and untraining.SSVEP can be divided into low-frequency SSVEP,intermediate-frequency SSVEP and high-frequency SSVEP according to the flicker frequency of its stimulus source.Compared with BCI based on high-frequency SSVEP,BCI based on low-frequency SSVEP has disadvantages such as prone to induce fatigue and photosensitive epilepsy.Therefore,BCI research based on high-frequency SSVEP may have high practical value and outstanding prospects in the future.This article focuses on the research of BCI based on high-frequency SSVEP and carries out the following research work:Considering that high-frequency SSVEP has lower response intensity than low-frequency and intermediate-frequency SSVEP,it is greatly affected by EMG artifacts and has low recognition rate.Empirical modal decomposition(EMD)is used to combine with canonical correlation analysis(CCA)and multivariate synchronization index(MSI),respectively.EMD-CCA method and EMD-MSI method were proposed for feature extraction and recognition of high-frequency SSVEP.Both methods use EMD to analyze the SSVEP signal before recognizing the frequency of the signal to obtain multiple Intrinsic Mode Functions(IMF),calculate the frequency band energy of each IMF component near the stimulation frequency band,and select the IMF component containing information related to the stimulus frequency as the feature matrix component.EMD-CCA uses the CCA method to analyze the feature matrix components and obtain the typical correlation coefficient.Based on this,SSVEP EEG signals produced by different visual stimulitions are identified and classified.EMD-MSI uses the MSI method to calculate the maximum synchronization index,and then classifies SSVEP signals.Compared with the traditional method of directly recognizing EEG signals using CCA or MSI,the two methods proposed in this paper extract feature components containing high-frequency stimulation information from EEG signals by adding the steps of EMD decomposition and IMF component selection.These two methods reduce the impact of artifact components contained in the signal and improve the recognition rate of high-frequency SSVEP.This paper designed and implemented an offline experiment,collected the two-class and four-class high-frequency SSVEP EEG signals,and verified the EMD-CCA method and EMD-MSI method.The experimental results showed that EMD-CCA and EMD-MSI methods could effectively identify high-frequency SSVEP signals and significantly improve the classification rate of high-frequency SSVEP.Finally,based on the above algorithms and simulation experiment researches,an online BCI system based on high-frequency SSVEP was implemented.Subjects can choose to look at the LED lights flashing at different frequencies to compare the size of two randomly presented playing cards.The system includes three parts: visual stimulator,EEG signal acquisition equipment and host computer.In the experiment,LED visual stimulator with adjustable blinking frequency was designed and produced;Neuroscan system was used to collect and transmit EEG signals in real time;LABVIEW is used to develop host computer software,process SSVEP signal,and display recognition results.In order to verify the performance of the system,high-frequency SSVEP pseudo-online experiments and online experiments were designed and implemented.Eight subjects participated in the pseudo-online experiments and online experiments.The experimental results showed that they could effectively manipulate the online BCI system to complete the playing card size comparison game.One of the participants has a single group online control accuracy rate of 100%.