A system for VEP detection and stimulus phase discrimination

The development of a system for detecting the presence and estimating the phase of the visual evoked potential in the electroencephalogram (EEG) of a human subject is presented. The stimulus set consists of several light emitting diodes (LEDs), all flashing at a rate of 10 Hz but with differing phases. The system employs a combination of wavelet prefiltering, harmonic preprocessing, neural network detection and phase discrimination and postprocessing of the neural network outputs. After an initial training period, the system is able to determine from the EEG at which of the LEDs the subject is looking with a rate of error of less than 20% within an interval of 0.8 seconds.