Estimation of driver attention using Visually Evoked Potentials

We propose a system for estimating driver attention levels using Visually evoked potentials (VEP), computed from the EEG signals of the visual cortex. We investigate the use of both steady state VEP (SSVEP) and pattern onset VEP (POVEP) for this purpose. The subject fixates on a flickering stimulus, generating a Steady State VEP (SSVEP). Occasionally, a random stimulus is flashed on the screen, and the subsequent POVEP is also analyzed. It is seen that the SSVEP is related to the attention levels of the subjects. The sudden stimulus also generates local maxima/minima values for the POVEP, at the P2 and N2 components. Entropy measures of the frequency response of both the responses could also be used to characterize the occurrence of stimuli. We also propose a system architecture for a driver alertness system, which fuses the above process and a vision based traffic analysis system, to alert the driver well in advance of any decrease in attention.