Chapter 1 Introduction 1 . 1 Brain - Computer Interfaces

OF THESIS ANALYSIS OF TEMPORAL STRUCTURE AND NORMALITY IN EEG DATA The thesis examines normality and temporal correlation of samples in EEG data. Presented in the context of Bayesian classification, empirical results quantify the advantages of modeling the temporal information. A connection to the theoretical background of the underlying classifiers is also discussed. Using a five-task six-channel dataset, the experiments demonstrate that an increase in performance can be observed by simple consideration of multiple samples. Exploitation of temporal structure leads to additional improvement, but not always. The measurement of Normality is used to demonstrate an alternative to cross-validation, where each class is considered independently of the others. Artem Sokolov Department of Computer Science Colorado State University Fort Collins, CO 80523 Spring 2007