Information Technology Reference
In-Depth Information
2
2
0 : 5
1if
ð ½ <f
z
½
n
g
þ ½ =f
z
½
n
g
Þ
r
\
d k Þ ¼
ð
12
Þ
0
otherwise
where, 1
N.Ifr CTM95 is the radius corresponding to 95 % CTM, then area of
analytic signal can be de
k
ned as:
A analytic ¼ p r CTM95
ð
13
Þ
2.3.2 Second-Order Difference Plot and Area Computation of Elliptical
Region
The second-order difference plot (SODP) provides a graph of successive rates
against each other and has been used to measure the variability present in EEG and
center of pressure (COP) signals (Thuraisingham et al. 2007 ; Pachori et al. 2009 ).
Useful diagnostic information can be extracted from SODP of the IMFs of EEG
signals. The area of SODP of IMFs of EEG signals can be used as features for
classi
cation of normal and epileptic seizure EEG signals. The SODP of signal x
½
n
½
½
can be obtained by plotting X
n
against Y
n
which are de
ned as (Cohen et al.
1996 ),
X
½
n
¼
x
½
n
þ
1
x
½
n
ð
14
Þ
Y
½
n
¼
x
½
n
þ
2
x
½
n
þ
1
ð
15
Þ
In SODP above mentioned successive rates are plotted against each other,
consequently provides rate of variability of data. The 95 % con
dence ellipse area
can be used to determine the con
dence area of SODP of IMFs which covers
around 95 % of the points. SODP corresponding to the normal and epileptic seizure
EEG signals and their
first four intrinsic mode functions are shown in Figs. 5 and 6 ,
respectively. These
of
different IMFs of EEG signals. The SODP of IMFs of EEG signals exhibit elliptical
patterns, the area of ellipse in SODP of IMFs has been used as a feature for
classi
figures represent trace of two successive rates, X
½
n
and Y
½
n
cation of epileptic seizure and seizure-free EEG signals (Pachori and Patidar
2014 ). In this work, we have used the area parameter computed from the SODP of
IMFs as a feature for classi
cation of normal and epilpetic seizure EEG signals. The
procedure to compute the 95 % con
dence ellipse area from the SODP can be given
as (Prieto et al. 1996 ; Cavalheiro et al. 2009 ):
The
l X and
l Y are mean values of lX
½
n
and Y
½
n
as de
ned in Prieto et al. ( 1996 ),
Cavalheiro et al. ( 2009 ) and
l XY can be de
ned as,
r
1
N X X
l XY ¼
½
n
Y
½
n
ð
16
Þ
Search WWH ::




Custom Search