During the operation of the visual stimulated brain-computer interfaces (BCIs), users suffer fatiguewhen gazing at the stimulus. Loss of attention and decreased vigilance caused by fatigue impair the users' cognitive capabilities, which consequently degrade the EEG quality and system performance. Cognitive neurodynamics measured by EEG activity are related to the generalized performance on cognitive tasks and mental efforts. Therefore, time varying EEG spectral analysis is proposed as an objective method to evaluate fatigue during the cognitive process in the visual stimulated BCIs. The results shows that the increase in theta, alpha and (theta + alpha)/beta, as well as the decrease in theta/alpha are associated with the increasing fatigue level during the cognitive process.