Single-Channel EEG Analysis - Quantitative EEG Analysis Methods and Clinical Applications - page 97

10

0

EEG following brain injury

− 10

0 2000 4000 6000 8000 10000 12000

(a)

0.5

0.4

Tsallis TDE ( w =64)

0.3

0 2000 4000 6000 8000 10000 12000

(b)

0.5

0.4

Tsallis TDE ( w =128)

0.3

0 2000 4000 6000 8000 10000 12000

(c)

0.5

0.4

Tsallis TDE ( w =256)

0.3

0 2000 4000 6000 8000 10000 12000

(d)

Figure 3.31 The role of window size in TDE: (a) 40-second EEG segment selected from the recovery
of brain asphyxia, which includes three typical spikes; and (b-d) TDE plots for different window size
( w = 64, 128, and 256 samples). The sliding step is set to one sample ( Δ = 1). The nonextensive

parameter q = 3.0. Partition number M = 10.

3.3.3.2 Window Lag
Because TDE is usually implemented with overlapping sliding windows, the win-
dow lag Δ defines the minimal time interval between two TDE values. Therefore, Δ

is actually the downsampling factor in TDE, where usually 1 by default. Figure

3.33 illustrates the influence of window lag on the TDE for the same EEG shown in
Figure 3.32. Comparing Figure 3.33(b-d), we see that Figure 3.33(c, d) actually
selected the TDE values in Figure 3.33(b) every other 64 or 128 samples,
respectively.

Δ=

3.3.3.3 Partitioning
One of the most important steps in TDE analysis is partitioning the signals to get the
probability distribution { P i }, particularly in histogram-based PDF estimation. The
three issues discussed next should be considered in partitioning.

Range of the Partitioning
To obtain the probability distribution { P i }, the EEG amplitudes should be parti-
tioned into a number ( M ) of bins. By default, some toolboxes, such as MATLAB,
create the histogram binning according to the range of the EEG, that is, the maxi-
mum and minimum, of the signal. Obviously, such a partitioning is easily affected