|
ИСТИНА |
Войти в систему Регистрация |
ИПМех РАН |
||
P300 BCI speller with human face images: the effects of stimuli variabilityтезисы доклада
- Авторы: Kitaitcev Andrey, Ganin Ilya, Grigoryan Rafael, Vasilyev Anatoly, Kaplan Alexander
- Сборник: Proceedings of the 6th International Conference Brain-computer interface: Science & Practice. Samara 2020
- Тезисы
- Год издания: 2020
- Издательство: Самарский государственный экономический университет
- Местоположение издательства: Самара
- Первая страница: 18
- Последняя страница: 19
- Аннотация: P300 BCI is based on analysis of event-related potentials (ERP) elicited in response to attended target stimuli. Usually P300 BCI utilizes only two stimuli classes: rare targets and frequent non-targets. We decided to add some variability inside each stimulus class making some realizations of stimuli appear more frequent than another. Our hypothesis was that such variability can increase the amount of attentional resources for stimulus processing thus promoting the enhancing of ERP amplitudes and probably improving the classification accuracy.16 healthy young subjects went through the P300 BCI speller task in this study. The 6x6 matrix contained grey ellipse (‘base pattern’) in each stimulus position on the black background. The base patterns in rows or columns were replaced with human face images (happy or neutral) during stimulation without changing of brightness in stimulus position. Totally there were two conditions with variability - ‘Rare happy’ (25% of face images were happy, 75% - neutral) and ‘Rare neutral’ (25% - neutral, 75% - happy), and two conditions without variability – ‘All happy’ and ‘All neutral’.ERP analysis elicited strong P200 component exceeding the amplitude of P300 component, typical for P300 BCI. The variability of stimuli as well as emotionality influenced the ERP amplitudes: both P200 and P300 components increased for rare stimuli and were higher for happy face images than for neutrals. The absence of significant variability influence on BCI accuracy can be explained by using ordinary classification algorithms, that are not adapted to heterogenous classes with both rare and frequent stimuli. Nevertheless, the obtained results demonstrate the possibility of further using of stimuli with variability for improving efficiency of P300 BCI systems.
- Добавил в систему: Ганин Илья Петрович