Brain Computer Interface
Brain Computer Interfaces (BCIs) enable motor disabled and healthy persons to operate electrical devices and computers directly with their brain activity. Our BCI recognizes and classifies different brain activation patterns associated with real movements and movement attempts made by tetraiplegic persons. One of our aims is to examine whether subjects with no previous experience of BCIs could achieve satisfactory performance after a short training period. Figure 1 demonstrates a BCI in use.
Figure 1. The user has a EEG cap on. By thinking about left and right hand movement the user controls the virtual keyboard with her brain activity. Copyright © 2003 by LCE.It would be important to understand the signals used in BCI applications. We have concentrated on motor cortex activity. Like most other BCI groups, we measure the electric activity of the brain using electroencephalography (EEG). We have also examined the feasibility of magnetoencephalography, MEG (see Figure 2), for BCI use.
Figure 2. Subject is being prepared for MEG experiment. EEG is also recorded. Copyright © 2004 by LCE.Feedback plays an important role when learning to use a BCI. In BCI training, the most commonly used feedback modality is visual feedback. Visual attention, however, might be needed for application control: to drive a wheelchair, to observe the environment, etc. It would be important to also test other feedback modalities. We have started to do tests with haptic feedback. Figure 3 shows an experiment where the subject receives tactile stimuli to the lower neck while learning to control a robot wheelchair in a simulated environment.
Figure 3. Subject is learning to control a robot wheelchair simulator while receiving haptic feedback to the lower neck. Copyright © 2006 by LCE.More information on all our current research can be found in our publication list
We have developed an online EEG-based BCI system which uses MATLAB in data analysis. In our first online experiments healthy persons performed real finger movements. The subjects' task was to move a circle from the centre of the screen to a target on either side of the screen. Four subjects were able to control the BCI with a mean classification rate of 73.5 %. To make the learning process faster, we have started to train the classifier online. During training periods, the classifier is adapted to the user's brain activity after each trial in a supervised manner, i.e. information about correct class is used. Thus, subjects can start using BCI from the beginning of the experiment without a separate training session. The method was tested with 10 healthy subjects and 6 tetraplegics. Figure 4 shows one tetraplegic person using our BCI.
Figure 4. Tetraplegic patient attempts left or right hand movements and tries to move the circle from the middle of the screen to the target Copyright © 2005 by LCE.
The healthy subjects controlled the BCI with a mean accuracy of 74.4 %. The performance of the best subject was 95 %. Three out of six of the tetraplegic subjects learned to control the BCI above chance level. Note that they did not move their hands but attempted to move them. Our current classification method works better with healthy people performing real hand movements than with motor-disabled persons attempting hand movements. One of our present interests is to examine if classification could be done more often than every 2 s. We also aim to find out more robust signals and analysis methods so that most tetraplegics would learn the use of BCIs relatively quickly.
Funding and co-operation
The BCI group at HUT consists of M.Sc. Laura Kauhanen (née Laitinen), M.Sc. Pasi Jylänki and research student Tapio Palomäki. Other persons currently involved are Academy Fellow Harri Valpola and D.Sc. (Tech) Aki Vehtari. This research is led by Academy Professor Mikko Sams, Academy Professor Kimmo Kaski and Dr. Tech. Jukka Heikkonen. See more information about the group.
Contact on page related matters email@example.com
This page has been updated 2.10.2006.