Spatio-temporal modeling of neural source activation from EEG data

This paper proposes a new computer-vision based information visualization paradigm for the electrophysiological study of face recognition. The proposed approach first generates video sequences of voltage maps from EEG data. Next, projections of active sources are detected in each frame using colour information and spatiotemporal consistency. The evolution of source activation is thus translated into a deformable motion of 2D patterns. Hence, the last step of the proposed approach builds a new motion representation, called the Spatio-Temporal Activation Response (STAR), which extracts stimulus- and subject-specific information about neural source activations occurring during the experiment. It is shown that STAR is able to capture relevant information about differences in the cognitive representations elicited by two different visual stimuli.