Biomedical Engineering Reference
In-Depth Information
60
P
1
40
20
0
−
20
−
40
−
60
P
7
−
80
0
200
400
600
800
1000
(a)
60
P
1
40
20
P
1
0
−
20
P
7
−
40
−
60
P
7
−
80
0
200
400
600
800
1000
(b)
0.5
0.4
0.3
Fixed partitioning
0.2
Adaptive
partitioning
0.1
0
200
400
600
800
1000
(c)
Figure 3.35
Two approaches to partitioning: A 4-second baseline EEG was scaled to 0.35 original
amplitudes in its second half so that an evident amplitude change is clearly shown. Two approaches
to partitioning are applied: (a) fixed partitioning for all sliding windows (
M
7 in this case), and (b)
adaptive partitioning (
M
=
7) dependent on the amplitude distribution within each sliding window.
=
Number of Partitions
The partitions, or bins, correspond to the microstates in (3.61), (3.64), and (3.66).
To obtain a reliable probability distribution {
P
i
} for smaller windows (e.g.,
w
128),
we recommend a partitioning number of less than 10. When analyzing long-
term activity with large sliding windows (e.g.,
w
=
2,048), partitions could be up to
M
=
30.
3.3.3.4 Entropic Index
Before implementing the nonextensive entropy of (3.64), the entropic index
q
has to
be determined. The variable
q
represents the degree of nonextensivity of the system,
Search WWH ::
Custom Search