Design of an online BCI system based on CCA detection method

Recently, more and more SSVEP (steady-state visual evoked potential) based BCIs (brain-computer interfaces) are developed to control external devices such as robots and wheelchairs. There have been many different methods for detecting the presence of SSVEPs. In this paper, CCA (canonical correlation analysis) detection method and PSD (power spectral density) detection method are compared in offline experiments. Results show that CCA has a much better performance than PSD. Therefore, CCA detection method is used in the online SSVEP-based BCI system with three targets. Three subjects participated to control a virtual wheeled robot in SIGVerse simulation environment. All of the subjects were able to use this BCI system and achieving an average accuracy of 89.8%, 95.6%, 92.5% respectively.