The Study Of A Novel Hybrid BCI Paradigm Based On The Combination Of Multiple EEG Feature

Brain-computer interface (BCI) as a communication system provides a totally new approach to create direct connection between brain activity and external devices for the paralyzed patients to communicate with the environment. In the system, messages or commands that an individual sends to the external world do not pass through the brain’s normal output pathways of peripheral nerves and muscles. The EEG-based BCI has the advantages of low cost, easy to implement, and non-invasive, made it the main direction of current brain computer interface research.In the traditional EEG-based BCI, only a single EEG feature is used as the input signal, however, every single EEG signal has characteristics and limitations, so there are some defects and shortcomings for the traditional BCI. P300-Speller is one of the classic paradigm to discriminate target using P300signal feature as the input signal, in which the characters matrix flash its rows and columns sequentially to elicits P300signal. But there still exists some disadvantages in the system:numerous flash times of a flashing cycle, low information transfer rate, and low signal-noise ratio.In this dissertation, some improvements focusing on these aspects have been disscussed. On the basis of the traditional P300-Speller, the hybrid BCI paradigm based on multiple EEG features is proposed, using the parallel flash stimulation. Study is divided into two stages as follows:At first, the preliminary paradigm based on the combination of SSVEP and P300is proposed and a3by3character matrix is devised to present the paradigm. When a subject keeps his/her attention on one character, the characters in the same row highlighted (flash) by blue frame in a random order can elicit the P300and the characters in the same column flashing on and off modulated at the same frequency can elicit SSVEP. The desired character can be distinguished through respectively detecting the P300feature and the SSVEP feature. The preliminary classification analysis of P300feature and SSVEP feature have been achieved, the feasibility of the paradigm is verified. Compared with the traditional P300-Speller, the new paradigm model using different row/column stimulation, and the times of flashing all characters are reduced to half, finally the efficiency of flash stimulation is improvedTo develop more practical BCI paradigm and improve the number of the candidate character, the character matrix is enlarged to6by6. And6colors are used as a factor to enhance the stimulations. In offline analysis, the results indicate that when the subject stare at a character, the two components of ERP (ERPs, P300and N200) are elicited by the chromatic frame and the SSVEP is elicited by the flash at the constant frequency. In order to test the classification effect of the characteristic signals, the classification mode of support vector machine (SVM) is design to detect ERPs feature, and the characteristics of target frequency is analyzed to detect SSVEP feature in the frequency domain, and the row or column is distinguished respectively through detecting different evoked potentials (ERPs and SSVEP). When the two kinds of EEG feature are both detected correctly, then. according to the intersection of row and column, the desired character is recognized. The results of character recognition show that character recognition accuracy is better in the hybrid Speller than the traditional P300-Speller and the information transfer rate also increased significantly, up to35bit/min.These results indicate that compare to the traditional P300-Speller, the time flashing all characters of hybrid speller is reduced to half, which effectively economize the time efficiency of elicitation. Multiple evoked-potentials are elicited simultaneously in the new parallel flash mode, and multiple signal features are mixed together, which increase the amount of identifiable EEG features. And the advanced algorithms are used to extract and detect signal features, finally the information transfer rate is improved without reducing the accuracy of character recognition, thus, the hybrid paradigm provides greatly potential for expanding the disign of the BCI speller.