Brain-Computer Interface and Virtual Reality

The bioelectric brain activity recorded from the intact scalp by the electroencephalogram (EEG) can be modified by different types of mental activity (thoughts) without performing any physical movement or speech. The thought- related EEG changes can be transformed into a control signal, when the EEG is used as input to a Brain-Computer Interface (BCI). The BCI output signal can be used, for example, to help a patient to select mentally letters out of the alphabet and write words (Virtual Keyboard) or to control assistive technologies. Both applications can improve herewith the quality of life for people with severe physical disabilities, suffering, for example, from amyotrophic lateral sclerosis or muscular dystrophy. The novel conception of this project is to use such an EEG-based BCI together with different types of online visualization of dynamic brain activity as a feedback mechanism to attain control over the ongoing EEG. Such a feed-back training could be applied to enhance the biofeedback therapy in the rehabilitation of various neurological and psychological disorders. One specific application could be e.g., to reduce seizures in patients with epilepsy. Another novel aspect of this project is to combine Virtual Reality (VR) and BCI technology, because VR provides immensive and controllable experimental environments and is extremely suitable for generation of feed back. Within the project special effort is devoted to continuous EEG feature extraction, feature classification and the most important task of feature mapping and visualization. For the latter different approaches for object visualization and visualization spaces will be studied, whereby the mapping needs to be adaptive (i.e. learnable).

Project P16326-B02 was focused (i) to develop an uncued (self-paced), asynchronous BCI system based on analyzing the dynamics of brain oscillations, (ii) to the integration of VR and BCI technology and to (iii) study the effects of different types of visual feedback in BCI experiments. It has been demonstrated, that self-paced BCI navigation in a virtual environment is possible with a minimum number of EEG channels and optimized EEG features. The self-paced operation mode requires that the BCI is constantly analyzing and interpreting the EEG activity. Due to the non-stationarity and inherent variability of the EEG signal, a very flexible and adaptive real-time system had to be developed and implemented. This implied that several "state-of-the-art" classifiers, as well as feature mapping and optimization methods were analyzed with respect to minimizing the number of EEG sensors and maximizing the classification accuracy between different mental activities. An equally important finding was that kinesthetic motor imagery is more efficient than visual motor imagery. This allows giving more accurate instructions to the user and in this way to reduce the training time and improve motor- imagery-based BCI control. Experiments with Virtual Reality feedback showed that moving body-parts reveal stronger EEG activation compared to geometric objects. This provides further evidence for some extent of motor processing related to visual presentation of objects and implies a greater involvement of motor areas in the brain.