An adaptive BCI system for virtual navigation

A brain computer interface (BCI) builds an additional pathway between the brain and external devices. This paper introduces a BCI system combined with virtual reality (VR) technology that allows users to navigate in the virtual apartment by motor imagery. Since performance in online sessions is sometimes found to decrease due to non-stationarity of electroencephalogram (EEG) signals, we proposed a method based on adaptive probabilistic neural network (APNN) to overcome this phenomenon. Online experiments were carried out using the BCI system and offline analysis was conducted to make a comparison among different methods. Online results demonstrated that most subjects were able to perform well on the system, and offline analysis showed that the APNN was comparatively predominant.