Correlation dimension computation of EEG time series

Abstract

In this thesis, a software package, to be used in correlation dimension (D2) computation of human electroencephalograms (EEG), is developed. The main algorithm calculating the D2 is based on the Grassberger Procaccia (G-P) theorem. Rigorous algorithms are developed to speed up the process without a loss in accuracy. The package is developed under Windows 95 in Delphi 2.0 environment, which enables the program to be a natural part of the contemporary 32 bit environments, and provides the user with a user friendly graphical user interface. The software developed is applied to two groups of signals. 1) Signals, whose D2's are known a priori- a sinusoidal, a Henon map, and a segment of white noise. 2) EEG samples recorded (whose D2's have to be calculated from experimental measurements) under various experimental conditions, from various groups of individuals, which comprise steady state responses of flash driven subjects, alpha dominant waves, beta dominant waves, and beta dominant waves from subjects having minor epilepsy. The numerical results obtained are analyzed, and compared with those in literature.