Multivariate composite estimators: new ways to track signals with application to human cerebral potentials

In this paper, we propose new multivariate composite estimators (MCE) to track multiple channel potential waves of individual and grand averages of recorded potentials at the scalp from a number of trials in a group of subjects. The advantages of these estimators over simple averages used in the literature are that they have larger signal to noise ratios (SNR), and that they have taken into account variations and correlations of the recorded potentials at the electrode sites as well as differences among subjects. Multivariate techniques and composite concept are used for deriving the estimators. We also provide an application to human event-related potentials in a memory for faces experiment.