Research On Brain - Computer Interface Based On Steady - State Visual Evoked Potential

A brain-computer interface (BCI) is a direct communication method between brain and machine without any use of peripheral nerves and muscles. Steady-state visual evoked potential (SSVEP)-based BCI systems have become a hot topic in BCI researches, for they have higher information transfer rates and rely on shorter calibration times.In this research, we focus on the optimization of SSVEP feature extraction and classification algorithms. The main results achieved in this research are listed as bellow:(1) We proposed a double partial least squares (D-PLS) model for SSVEP feature extraction and classification without calibration data. In our proposed D-PLS model, firstly, PLS was adopted as a spatial filter for extracting SSVEP feature from multi-channel EEG data, and then PLS was used to classify the extracted SSVEP signals.To validate the proposed D-PLS model, the performances of the PLS spatial filter and the PLS classifier were tested respectively. The experimental results confirmed that PLS can obtain high performance both on SSVEP feature extraction and classification, and the computing velocity of BCI system was improved obviously. On average, the classification accuracy can be improved 2～4% than the existing methods at various experimental conditions. Thus D-PLS is a promising way of enhancing the performance of SSVEP-based BCI systems.(2) We proposed a combination analysis of power spectral density analysis (PSDA), canonical correlation analysis (CCA), and D-PLS to recognize the stimulus frequency for SSVEP-based BCI. The experimental results confirmed that the combination analysis can improve the classification accuracy 2～5% than D-PLS averagely within shorter TW lengths (1.5～3.5 s).