Brain Computer Interface Research Based On P300and SSVEP

A Brain computer interface (BCI) is a new bridge which connects human mind with outside world. It would be beneficial for people with severely impaired motor systems, who could not communicate through physical activities.The most commonly used signals in EEG-based BCI systems are steady-state visual evoked potentials (SSVEPs) and P300potentials. BCIs based on SSVEPs can be used without training, which can obtain high bit rate (i.e. information transfer rate, ITR). P300is a kind of event related potentials which could be evoked by the oddball paradigm. Compared to other non-invasive BCIs, SSVEP and P300BCIs can obtain high bit rate and high accuracy. Although P300BCIs and SSVEP BCIs perform good in accuracy and information transfer rate, they still need to be improved further for practical application.The motivation of this paper is to research on experimental paradigm and classification methods. The main contributions of this paper are as follows:(1) A directional control system based on P300brain computer interface was designed. In the experiment, subjects were required to move a ball on the computer screen to the target position using P300BCI system. This system verified the effectiveness of P300BCI system, which could be used for further practical device.(2) A directional control system based on P300BCIs and SSVEP BCIs respectively was designed. The aim was to show the drawbacks and advantages of these two BCIs, when they were used in directional control task. The online experimental results showed that P300BCIs was more robust, when it was compared to SSVEP BCIs.(3) In order to improve the accuracy and bit rate, the methods of electrode selection were studied. Three methods (Sequential floating forward selection (SFFS), discrete particle swarm optimization (DPSO) and F-score) were used to select the optimal electrode channels. The performance of electrode channels obtained by SFFS, DPSO and F-score were compared with that of traditional electrode channels (i.e.,01,02and Oz), which were usually used in SSVEP BCI. The results show that SFFS and DPSO can obtain higher classification accuracy than traditional approach. Channel selection can not only reduce features for data analysis but also reduce the time for channel installation.