Characterization of evoked and induced activity in EEG and assessment of intertrial variability

Brain response to an internal or external event, is composed by the superposition of evoked and induced oscillatory activity, which reflect different brain mechanisms involved. Identification of such activations could serve for diagnostic purposes and provide useful tools for brain computer interfaces through insight on the activation of different brain regions. In this paper we study several statistical measures that have been proposed for identifying the nature of the involved activations. All these measures are based on some mean of an appropriate signal attribute over trials in the time/ frequency domain and do not characterize the variability across trials. In order to quantify trial-to-trial variability we consider a measure based on entropy, characterizing the distribution of power across trials. The results indicate that brain activations can be characterized and differentiated by their behavior from trial to trial.