SSVEP Based Brain-Computer Interface for Robot Control

A brain computer interface (BC!) using steady state visual evoked potentials (SSVEP) is presented. EEG was derived from 3 subjects to test the suitability of SSVEPs for robot control. To calculate features and to classify the EEG data Minimum Energy and Fast Fourier Transformation (FFT) with linear discriminant analysis (LDA) were used. Finally the change rate (fluctuation of the classification result) and the majority weight of the analysis algorithms were calculated to increase the robustness and to provide a zero-class classification. The implementation was tested with a robot that was able to move forward, backward, to the left and to the right and to stop. A high accuracy was achieved for all commands. Of special interest is that the robot stopped with high reliability if the subject did not watch at the stimulation LEDs and therefore successfully zero-class recognition was implemented.