抄録
EEG related to time dependent varying tasks are signals of non-stational processes, and ordinal information criteria for which stational processes are assumed can not be applied directly. We have proposed an analysis method for EEG signals related to time dependent varying tasks using an information criterion (WMDL) which has been published by one of the authors. By using the method, time dependent varying information amount of a time dependent varying EEG can be evaluated. Tasks performed in this research (II) were observing line drawings of a rounded spiral and a squared spiral whose drawing total time was 17seconds for each. The images of the line drawings were displayed on a CRT display. The subjects were seven graduate students. For measurements of EEG at observations of the images, an electrode cap with 19 electrodes was mounted on the subjects. For two electrodes related to motion activities, the information amount analysis was performed. As the result, on the time axis, most of the information amounts of the EEG signal of the rounded spiral is larger than that of the squared spiral for all seven subjects. For comparison, the information amounts of the primitive images of the line drawings using a Markov model were calculated theoretically and on the time axis, most of the information amounts of the rounded spiral are larger than that of the squared spiral whose information amount is large only at the corners. The important result for the tasks is that the inequality relation between the information amounts of the primitive images is preserved also in the brain signals. [J Physiol Sci. 2008;58 Suppl:S146]
