Thesis in Portuguese

Sleep Spindles, Decomposition, Parameterisation and Applications, 2010-FCT

(Fusos de sono: decomposição, parametrização e aplicações)

Abstract:

This thesis presents a study of sleep spindles. The sleep spindles are one of the
components of the electroencephalogram (EEG), characterized by a center frequency
in the range of 11-15 Hz and may have relevant clinical significance in the case of
neurodegenerative diseases.
The sleep spindles are isolated from the electroencephalogram (EEG) signal, by the
method of empirical mode decomposition (EMD) and can be viewed as
amplitude/frequency modulated (AM/FM) signals.
The demodulation of sleep spindles is carried out in two steps. In the first step,
waveforms of the instantaneous envelope (IE) and the instantaneous frequency (IF)
are estimated, through the method of demodulation by interpolation and AR
modulation. In the second step the IE and IF waveforms are approximated by an
AM/FM model and their parameters are calculated.
In this thesis were adopted three AM/FM models, in  order to compare them and
choose the most appropriate for the study signals.  In the calculation of model
parameters were also used two different methods, namely, the method of least
squares and genetic algorithms.  Tests of the model are made based on real sleep
spindles.
Finally, the parameters of the AM/FM model are used to compare from healthy
subjects and individuals with Alzheimer's.
Keywords:  Sleep spindles; Empirical mode decomposition; Intrinsic mode function;
AM/FM model; Least square; Genetic algorithms; Dementia.

Thesis PDF link: http://run.unl.pt/bitstream/10362/4108/1/Almeida_1010.pdf