Evolutionary Factor Analysis of EEG Data

• Activity Number: 15
• Type: Topic Contributed
• Date/Time: Sunday, August 1, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Giovanni Motta*+ and Hernando Ombao
• Companies: Maastricht University and Brown University
• Keywords: Electroencephalography ; Local Stationarity ; Principal Components ; Source Localization

Abstract

We model and estimate the dynamic structure of multi-channel electroencephalograms (EEG) in a motor-visual experiment. Both the variance of each channel and cross-covariance between a pair of channels evolve over time. Moreover, the cross-covariance profiles display a common structure across all pairs. Based on these observations, we apply the Evolutionary Factor Analysis to our EEG data. We explain the common co-movements of EEG signals through the existence of three factors. These latent factors are primarily responsible for processing the visual-motor task which, through the loadings, drives the behavior of the signals observed at different channels. We show that the loadings are slowly varying over time and that, the higher the weight on one factor the smaller the weight on the other two factors. This reflects: smooth evolution of the dynamics and orthogonality of the latent sources.