Recognition of mental activity type with EEG

A new method for recognition of mental activity type with EEG is elaborated. The recognition is based on the analysis of EEG rhythms. Few (three-five) seconds of multi-channel EEG record is enough to accomplish the recognition since an artificial neural network (ANN) is utilized for the purpose. Best results are attained when using the smoothed power spectra of single EEG trials as an input to ANN, i.e. when the classification is based on the analysis of the frequency and the topography of rhythms. At the first stage, the ANN learns to detect mental activity types correctly using the learning set of data, for which the activity type is assumed to be known for every portion of EEG. Later, the network recognizes the unknown EEG data. To induce the thoughts of certain kinds, the appropriate tasks were prepared and presented to subjects on a computer monitor. The tasks belonged to two principally different cognitive styles, i.e. verbal and spatial. For each of these two, four different task types were designed, and for each of those 100 actual tasks were prepared. So, a total of 800 tasks were presented to every subject. There were an overall of twenty adult healthy subjects. Our results show that reliable discrimination between principal cognitive styles can be easily done on base of short EEG trials. For instance, when teaching the ANN to distinguish between basic thinking modes (verbal vs. spatial) with four task types (two verbal and two spatial), the other four are ascribed to the appropriate basic classes correctly. The average (over all subjects and task types) recognition score is about 90% in this case. The discrimination of particular task types within one basic class is possible but with lower scores (66%). Basing on literature data, we hypothesize that brain processes such as directional attention, memory load, and motivational trend may underlay the appearance of distinctive rhythmical patterns. A mu-rhythm may also play a significant role since execution of tasks might be accompanied by involuntary subtle movements and virtual speech production or, vice-versa, their suppression. The observed rhythmical patterns reveal to be highly specific, both to a subject and to a certain mental activity type. At the same time, they are surprisingly stable in time and are preserved for at least several months. Thus, a set of characteristic EEG rhythms pertinent to certain mental activity types forms an individual subject's ‘EEG portrait’.