Simon Porras is presenting his Master thesis,


Title: Time-Frequency Multitaper Analysis of EEG from children with ADHD

Abstract:

The main subject of this thesis is to apply different time-frequency
tools to Electroencephalogram (EEG) data taken from children with
Attention-Deficit/Hyperactivity Disorder (ADHD). The object is to find
differences and/or similarities in the EEG frequency content of ADHD
group and Control Group. To this purpose, three different spectral
estimation methods are studied and applied to the EEG data. The thesis
begins with an overview of important concepts such as the Fourier
transform as a basis of development for the rest of the work. First,
the signal properties are defined, such as sample rate, number of
samples and sampling time. Then the EEG procedure is explained and the
process of obtaining the EEG data is also described. The Stroop Test
was used to record the EEG data, it is a measure of the effect of
interference on performance of a colour identification task. We then
explain the Short-Time Fourier Transform (STFT), and two multiple
window methods that can be used in this context: Thomson's Multiple
Window method and the Peak Matched Multiple Windows (PM MW) for
spectral estimation. Thomson's Multiple Window method designs tapers
which aim to suppress side-lobe leakage, while giving stable estimates
with the use of multitapers. The PM MW method reduces bias at peak
frequencies by using matching windows, and it lowers variance with the
average of uncorrelated periodograms. With the methods used, high
energy levels of delta and theta frequencies (0-8 Hz) were found to be
present in the EEGs of both groups, during the instants when stimuli
were presented. Both the use of PM MW and Thomson method give a better
performance in the time-frequency domain than the usage of a single
Hanning window.