Virtual Reality Embodiment in Motor Imagery Brain–Computer Interface Training

Abstract

This study investigates how the avatar embodiment in virtual reality (VR) influences training for operation of motor imagery brain–computer interfaces (MI-BCIs). In previous studies, we demonstrated the possibility to induce a BCI-based VR avatar embodiment (using mental commands instead of motion tracking) for purposes of training facilitation. This paper examines the relationship between BCI performance and subjective levels of embodiment, as well as differences in BCI performance achieved with training in two different VR environments. MI-BCI training variants with the following forms of feedback were studied: a) with feedback exploiting avatar embodiment in immersive VR (avatar movements indicating the user performance), b) with feedback exploiting avatar embodiment in gamified and progressively paced VR training, and c) with symbolic feedback in the standard training with Graz protocol. On-line performance from the BCI experiments and questionnaires on the sense of ownership and sense of agency toward the virtual avatar were studied. Questionnaire analysis showed that a robust sense of embodiment arose in the VR training environments, with a strong correlation between the reported ownership and agency toward the avatar. Interestingly, the achieved BCI performance was uncorrelated with both the ownership and the agency. Using gamification further increased the performance (but not the reported sense of ownership) in the training session. Embodiment in VR mediated by synchrony between mental commands and visual stimulation in VR arose under different conditions than embodiment based on visuo-motor synchrony. Consistency between the perceived sense of ownership and agency plays a more important role than the ability to issue MI-BCI commands correctly. These findings help to elucidate the positive effect of embodiment to initial steps in BCI training and can be leveraged in future MI-BCI training designs.