A P300 BCI Signal Detection Algorithm Based On Ensemble Of SVM

Brain-Computer Interface(BCI) is a kind of direct way of information interchange between human brain and the computer or the other electronic equipment. It is a kind of brand-new information interchance systerm not relying on the normal output channel of brain. BCI is an inevitable result with the development of electronic and computer technology. As a wholly new information communication and control technology, BCI is winning more and more attention. BCI may not only provide the paralyzed with an effective communication and control channels with outside world, but also can be used in the other fields such as spaceflight, special circumstance working and so on.Many studies on neurology science show that brain can produce special electrical activities when the subject is presented with task-relevant stimuli. According to these electrical activities the thoughts that the subject wants to express can be recognized, as long as translating the thoughts into device control command. Hence, the key of BCI system is how to translate electrical activities (P300 signals) recorded by Electroencephalograph into device control commands quickly and accurately.Brain-Computer Interface P300 speller aims at helping patients unable to ctivate muscles to spell words by means of their brain signal activities. Associated to this BCI paradigm, there is the problem of classifying electroencephalogram signals related to responses to some visual stimulation. This paper addresses the problem of signal responses variability within a single subject in such Brain-Computer Interface. We propose several method that copes with feature extraction, classify and character identification. The algorithm based on the characteristic parameters of the direct wave uses characteristic wave in time field to compute the parameters,which fully describes the wave in time field of P300; The algorithm based on the wavelet transition and PCA can locate the P300 part in the samples and merge these features. For the design of classify, we propose a method that copes with such variabilities through an ensemble of classifiers approach .Each classifier is composed of a linear Support Vector Machine trained on a small part of the available data and for which a channel selection procedure has been performed. Performances of our algorithm have been evaluated on dataset II of the BCI Competition III and has yielded the best performance of the competition.and provide that the method is effective via the data we get from the Nueroscan system.