2007 Volume 12 Issue 1 Pages 11-16
In this article, we propose an innovative technology of "Brain CAVE Interface". The electroencephalogram signals of steady-state visual evoked potentials, induced by virtual objects illuminated with constant frequencies, were recorded. Several pattern recognition techniques were investigated to discriminate between brain signals induced by one virtual object and that by another. Machine learning made a remarkable performance. The support vector machine with single trial data for 2.0 (1.0) seconds resulted in 94.6 (91.2) % of averaged recognition rate. This extended concept of non-invasive brain computer interface would enable us to realize intuitive and non-muscular manipulations of virtual objects. The promising visual evoked potentials in CAVE will be suggested in the viewpoint of direction controlling for walking application.
