ИСТИНА |
Войти в систему Регистрация |
|
ИСТИНА ИНХС РАН |
||
The current study was designed to determine whether the existence of personal context of tested information could be detected in one of the most commonly investigated and replicated ERP procedures, the oddball paradigm, realized in brain-computer interface (BCI). The visual P300 BCI, a well-known system utilizes the P300 event-related potential to select letters, icons or images arranged in a flicker matrix [Farwel, Donchin 1988]. In our study the 9 facial images were embedded in 3x3 matrixes and flashed in a random order during 160 ms each. The pause between two consecutive flashes was 80 ms. In our experiments, 6 healthy subjects were instructed to focus on a screen with 9 different facial images without special interest to any of them (paradigm 1) or silently count how many times one of the facial images was flashed (classical paradigm 2) and the same condition as paradigm 1 but one of face images was replaced by photo of tested subject (paradigm 3). In a training stage, 9 strange facial images were shown and each target was flashed 6 times with silently counting. In a run of on-line test stages the were shown 8 strange faces and 1 facial image of tested subject. Each image cell size was 5x5 cm on the screen. We repeated 5 runs for each image. Subjects were seated in a comfortable chair placed 90 cm from a 19 inch LCD screen. We recorded EEG data and computed on-line classification accuracies. EEG signals were recorded at Cz, Pz, O1, O2, Oz, PO7 and PO8 site following the 10-20 international system and each channel data was amplified by Enobio (Neuroelectrics, Spain). It was analyzed 800 ms from each start time of a stimulus, subtracting baseline calculated by using 150 ms pre-stimulus data. A linear discriminant analysis (Fisher discriminant) was used for classification of ERP from target and no target stimuli. A classification accuracy was calculated as (# correct trials) / (# total trials) *100%. It was shown that classification accuracies of classical paradigm 2 with focused attention to target stimulus were over 85% for all subjects while in paradigm 2 without attention the classification accuracies varied from stimulus to stimulus between 8 and 23% (11% chance level). The main result was found in paradigm 3 with no instruction for selective attention to any of images. In this condition the triggering of P300 BCI by the images of own face was detected with unexpected high accuracies for condition "without instruction to selective attention" - over 70% for all subjects. Thus, P300 BCI may be used to detect the existence of personal context in relation to tested information. It can be used to create automated systems for neurosemantics and neurolinguistic testing.