Recognition Method Of SSVEP Based On FSK Modulated Stimuli

Brain computer interface(BCI)technology realizes the communication between human brain and external devices.Because it does not need to rely on muscle tissue to control external devices,the technology brings hope to many disabled people.At present,a large number of teams at home and abroad have studied the technology,which has a wide application prospect.Among them,brain computer interface system based on steady state visual evoked potential(SSVEP)has gradually become the mainstream of brain computer interface system because of its simple experimental paradigm and high accuracy.Due to the influence of physiological mechanism and hardware technology,the existing SSVEP-BCI system has the problem of less instructions.To solve this problem,this paper adopts an experimental paradigm based on frequency shift keying(FSK)to increase the number of instructions in BCI system.Under the SSVEP-BCI system with this paradigm,this paper mainly studies the following aspectsIn the design of experimental paradigm: firstly,the relevant principles of FSK and SSVEP are analyzed.In the experiment,different parameters(stimulus frequency,unit symbol time length)are set to explore its impact on the system performance.Secondly,in view of the fact that visual stimuli have individual differences to subjects,a single frequency experiment is set up to compare the response of each subject to visual stimuli,and the stronger visual stimuli are selected to be used in the subsequent FSK experimental paradigm to increase the effectiveness of the experiment.Finally,the EEG data acquisition of several subjects was completed through the experiment.In signal preprocessing: for the problem of low signal-to-noise ratio of EEG signal,this paper proposes a preprocessing method of EEG signal based on multi-synchronous compression transform(MSST).Based on the simulation signal and real EEG signal,the method is used to preprocess the signal.Then,compared with the commonly used EEG preprocessing methods(empirical mode decomposition and singular value decomposition),it is found that MSST has better signal processing effect and can effectively improve the signal-to-noise ratio of SSVEP.In the recognition algorithm: in the BCI system based on FSK-SSVEP,for the problem that the local time feature is ignored by canonical correlation analysis,a recognition method of principal component analysis combined with coherent demodulation is proposed.Finally,a variety of preprocessing algorithms combined with different recognition algorithms are used to realize the recognition of instruction target.The results show that high accuracy(90.31%)can be obtained by using 4MSST and PCA-coherent demodulation.At the same time,the system has good performance(18.8bit/min)when 1.2s is selected as the unit symbol time.Therefore,it is proved that our method has a good application prospect in BCI system based on FSK-SSVEP.