Optimizing Visual Cues for Brain-Computer Interfaces

Abstract

Brain-Computer Interfaces (BCI) based on event related potentials (ERP) have been successfully developed for applications like virtual spellers and navigation systems. This study tests the use of visual stimuli unbalanced in the subject’s field of view to cue movement imagery tasks (left vs. right hand movement) and elicit early transient visual ERPs. These cues were compared with balanced cues on classification accuracy. Subject-specific ERP spatial filters were calculated for optimal group separation. The unbalanced cues improved movement imagery discrimination achieving errors as low as 3%. The unbalanced cues appear to enhance visual ERPs related to primary visual processing that reduce the classification error to a minimum soon after the cue presentation. This work suggests that the use of such visual interfaces may be of interest not only for the overall performance improvement of BCI application but also for its use on validation of control decisions.