Brain–Computer Interfaces - Index of

70 C. Neuper and G. Pfurtscheller
Fig. 1 Band power (11–13 Hz) time courses ±95% confidence interval displaying ERD and ERS
from training session without feedback (left) and session with feedback (right). These data are
from one able-bodied subject while he imagined left and right hand movement. Grey areas indicate
the time of cue presentation. Sites C3 and C4 are located over the left and right sensorimotor areas
of the brain, respectively
When a naïve user starts to practice hand motor imagery, a contralaterally dominant
desynchronization pattern is generally found. Changes in relevant EEG patterns
usually occur if the user is trained via feedback about the mental task performed.
Figure 1 shows an example of band power time courses of 11–13 Hz EEG activity
of one subject obtained at two times: during the initial screening without feedback;
and during a subsequent training session while a continuous feedback signal (moving
bar) was presented. The ERD/ERS curves differ during right versus left motor
imagery, with a significant band power decrease (ERD) over the contralateral (opposite)
hand area, and a band power increase (ERS) over the ipsilateral (same) side.
The feedback enhanced the difference between both patterns and therewith the classification
accuracy (see also [40]). The enhancement of oscillatory EEG activity
(ERS) during motor imagery is very important in BCI research, since larger ERS
leads to more accurate classification of single EEG trials.
BCI training can be run like a computer game to make participants more engaged
and motivated. In the “basket-game” paradigm, for example, the user has to mentally
move a falling ball into the correct goal (“basket”) marked on the screen ([41], see
also Fig. 2, left side). If the ball hits the correct basket, it is highlighted and points
are earned. The horizontal position of the ball is controlled via the BCI output signal,
and the ball’s velocity can be adjusted by the investigator.
Four male volunteers with spinal-cord injuries participated in a study using this
paradigm. None of them had any prior experience with BCI. Two bipolar EEG signals
were recorded from electrode positions close to C3 and C4, respectively. Two
different types of motor imagery (either right vs. left hand motor imagery or hand vs.
foot motor imagery) were used, and band power within the alpha band and the beta
band were classified. Based on each subject’s screening data, the best motor imagery
tasks were selected, and the classifier output (position of the ball) was weighted to
adjust the mean deflection to the middle of the target basket. This way, the BCI