Cognitive task classification based upon topographic EEG data.

EEG from 19 electrodes was used to classify which of 14 tasks each of seven subjects had performed. Stepwise discriminant analysis (SWDA) was used to classify the tasks based upon training on one half of the spectrally analyzed 1 min of data. Eighty six percent correct classification was achieved using principle components analysis (PCA) to determine the EEG bands to be used by the SWDA. Other approaches to deriving the EEG bands met with lower levels of success. The results indicate that frequency and topographical information about the EEG provides useful knowledge with regard to the nature of cognitive activity. Higher frequencies provided much of the information used by the classifier. The utility of this approach is discussed with regard to evaluating operator state in the work environment.