A robust brain computer interface with in situ stimuli in a stereoscopic head mounted display.

Authors: 
Blom K.J., González-Franco M., de la Vega J., Hintermüller C., Guger C. & Slater M.
Date of publication: 
2012
Journal: 
8th FENS Forum of Neuroscience

Brain-Computer Interfaces (BCI) that support interactions without physical exertion hold the promise of enabling those who cannot perform motor-based interactions new found freedoms. Prior research has shown that motor imagery approaches in projection-based immersive virtual reality systems are feasible. However, the extensive training required and low accuracy limit the applicability of this technique. We propose the steady state visually evoked potential (SSVEP) paradigm as a robust alternative which can achieve high success even under difficult conditions. We present our experience in applying SSVEP in a stereoscopic head mounted display (HMD). Up to five SSVEP stimuli are presented in situ as “overlaid” objects in the 3D virtual environment. During development, we used a variety of display configurations of various sizes and orientations of square colored stimuli. Since the HMD requires a 60Hz refresh rate, frequencies of 8.57, 10, 12, and 15Hz were used in a standard setup. We were able to achieve classifications rates (95%) close to those seen in desktop versions with at most two four minute training sessions. This setup proved robust in testing, using different training and testing/free-run configurations, including training on desktop and testing in the HMD and removal of the HMD between training and testing. Only slight degradations in the classification were experienced. Testing embedded in a robotic application has shown its feasibility under realistic conditions. The participant controls a small humanoid robot's movement via SSVEP controls. A stereoscopic viewpoint, via streaming camera images from the robot, forms a semi-controlled moving 3D environment. Three healthy users tested this setup. All achieved classification error rates below 5% during training. They reported positive control of robot movement and demonstrated the ability to control forward, left/right turning and idle states. This work is funded by the EU FET VERE project and the ERC Senior Project TRAVERSE.