Spatial-frequency Fusion Of MSFA For SSVEP Recognition And Its Application

Brain computer interface(BCI)opens up a direct channel for information exchange between the brains and the external environments,which has great significance of improving people's quality of life.Stable state visual evoked potential(SSVEP)has become a hot topic in BCI field due to its high information transmission rate,ease of use and short training time.In recent years,many efficient classification algorithms based on SSVEP have been proposed,but the communication speed is still difficult to meet the needs of online SSVEP-BCI systems;Lack of training data is another obstacle to the promotion of system.To solve these two problems,this thesis proposes an efficient SSVEP signal frequency recognition method,and designs a friendly online character input system.The specific work is as follows:1)This thesis proposes a new SSVEP frequency recognition method with spatialfrequency fusion strategy.Maximum signal fraction analysis(MSFA)is an efficient SSVEP frequency identification method.However,standard MSFA only utilizes the spatial filter corresponding to the maximum eigenvalue for multichannel SSVEP signals denoising and frequency identification,and discards other feature vectors containing discriminant information.Fo MSFA employs the information of all spatial filters from standard MSFA,and uses a nonlinear weighting function to fuse multiple sets of correlation coefficients to identify the frequency of SSVEP signal.Numerical results of two benchmark datasets and our laboratory dataset with 10 subjects show that Fo MSFA outperforms the standard MSFA and CCA-based methods.The spatial-frequency fusion strategy can effectively improve the performance of the MSFA algorithm,which has great significance of other multi-filter generating algorithms.2)This thesis implements an efficient online character input system that requires only a small amount of training data.Transfer learning technology is usually used to solve the problem of insufficient data in machine learning.This thesis combines the Fo MSFA method and transfer learning technology to solve the problem of low information transmission rate and insufficient training data in the online SSVEP-BCI system.The system uses Actic Hamp amplifier for data recording and processing;uses the least square transformation technology for cross subject transfer learning on MATLAB;and uses Fo MSFA method and multi-frequency learning technology for frequency recognition.The results showed that the subjects could input a single character quickly and accurately within 1.96 s.15 subjects performed character input with cues,the average accuracy rate and information transmission rate were 79.3% and161.9 bits/minute,respectively;10 subjects performed character input without cues,the average accuracy rate and information transmission rate were 80.0% and 163.5 bit/min,respectively.The combination of transfer learning technology and efficient classification algorithm provides a new idea for the development of the friendly SSVEP-BCI system.