<s>
EEG	B-Algorithm
analysis	I-Algorithm
is	O
exploiting	O
mathematical	O
signal	O
analysis	O
methods	O
and	O
computer	O
technology	O
to	O
extract	O
information	O
from	O
electroencephalography	B-Application
(	O
EEG	B-Application
)	O
signals	O
.	O
</s>
<s>
The	O
targets	O
of	O
EEG	B-Algorithm
analysis	I-Algorithm
are	O
to	O
help	O
researchers	O
gain	O
a	O
better	O
understanding	O
of	O
the	O
brain	O
;	O
assist	O
physicians	O
in	O
diagnosis	O
and	O
treatment	O
choices	O
;	O
and	O
to	O
boost	O
brain-computer	B-Application
interface	I-Application
(	O
BCI	O
)	O
technology	O
.	O
</s>
<s>
There	O
are	O
many	O
ways	O
to	O
roughly	O
categorize	O
EEG	B-Algorithm
analysis	I-Algorithm
methods	O
.	O
</s>
<s>
If	O
a	O
mathematical	O
model	O
is	O
exploited	O
to	O
fit	O
the	O
sampled	O
EEG	B-Application
signals	O
,	O
the	O
method	O
can	O
be	O
categorized	O
as	O
parametric	B-General_Concept
,	O
otherwise	O
,	O
it	O
is	O
a	O
non-parametric	O
method	O
.	O
</s>
<s>
Traditionally	O
,	O
most	O
EEG	B-Algorithm
analysis	I-Algorithm
methods	O
fall	O
into	O
four	O
categories	O
:	O
time	O
domain	O
,	O
frequency	O
domain	O
,	O
time-frequency	B-Algorithm
domain	I-Algorithm
,	O
and	O
nonlinear	O
methods	O
.	O
</s>
<s>
There	O
are	O
also	O
later	O
methods	O
including	O
deep	B-Algorithm
neural	I-Algorithm
networks	I-Algorithm
(	O
DNNs	O
)	O
.	O
</s>
<s>
Frequency	O
domain	O
analysis	O
,	O
also	O
known	O
as	O
spectral	O
analysis	O
,	O
is	O
the	O
most	O
conventional	O
yet	O
one	O
of	O
the	O
most	O
powerful	O
and	O
standard	O
methods	O
for	O
EEG	B-Algorithm
analysis	I-Algorithm
.	O
</s>
<s>
It	O
gives	O
insight	O
into	O
information	O
contained	O
in	O
the	O
frequency	O
domain	O
of	O
EEG	B-Application
waveforms	O
by	O
adopting	O
statistical	O
and	O
Fourier	B-Algorithm
Transform	I-Algorithm
methods	O
.	O
</s>
<s>
Among	O
all	O
the	O
spectral	O
methods	O
,	O
power	O
spectral	O
analysis	O
is	O
the	O
most	O
commonly	O
used	O
,	O
since	O
the	O
power	B-Algorithm
spectrum	I-Algorithm
reflects	O
the	O
'	O
frequency	O
content	O
 '	O
of	O
the	O
signal	O
or	O
the	O
distribution	O
of	O
signal	O
power	O
over	O
frequency	O
.	O
</s>
<s>
There	O
are	O
two	O
important	O
methods	O
for	O
time	O
domain	O
EEG	B-Algorithm
analysis	I-Algorithm
:	O
Linear	O
Prediction	O
and	O
Component	O
Analysis	O
.	O
</s>
<s>
Notably	O
,	O
the	O
parameters	O
in	O
time	O
domain	O
methods	O
are	O
entirely	O
based	O
on	O
time	O
,	O
but	O
they	O
can	O
also	O
be	O
extracted	O
from	O
statistical	O
moments	O
of	O
the	O
power	B-Algorithm
spectrum	I-Algorithm
.	O
</s>
<s>
Wavelet	B-Algorithm
Transform	I-Algorithm
,	O
a	O
typical	O
time-frequency	B-Algorithm
domain	I-Algorithm
method	O
,	O
can	O
extract	O
and	O
represent	O
properties	O
from	O
transient	O
biological	O
signals	O
.	O
</s>
<s>
Specifically	O
,	O
through	O
wavelet	O
decomposition	O
of	O
the	O
EEG	B-Application
records	O
,	O
transient	O
features	O
can	O
be	O
accurately	O
captured	O
and	O
localized	O
in	O
both	O
time	O
and	O
frequency	O
context	O
.	O
</s>
<s>
Thus	O
Wavelet	B-Algorithm
transform	I-Algorithm
is	O
like	O
a	O
mathematical	O
microscope	O
that	O
can	O
analyze	O
different	O
scales	O
of	O
neural	O
rhythms	O
and	O
investigate	O
small-scale	O
oscillations	O
of	O
the	O
brain	O
signals	O
while	O
ignoring	O
the	O
contribution	O
of	O
other	O
scales	O
.	O
</s>
<s>
Apart	O
from	O
Wavelet	B-Algorithm
Transform	I-Algorithm
,	O
there	O
is	O
another	O
prominent	O
time-frequency	O
method	O
called	O
Hilbert-Huang	B-Algorithm
Transform	I-Algorithm
,	O
which	O
can	O
decompose	O
EEG	B-Application
signals	O
into	O
a	O
set	O
of	O
oscillatory	O
components	O
called	O
Intrinsic	O
Mode	O
Function	O
(	O
IMF	O
)	O
in	O
order	O
to	O
capture	O
instantaneous	O
frequency	O
data	O
.	O
</s>
<s>
Many	O
phenomena	O
in	O
nature	O
are	O
nonlinear	O
and	O
non-stationary	O
,	O
and	O
so	O
are	O
EEG	B-Application
signals	O
.	O
</s>
<s>
This	O
attribute	O
adds	O
more	O
complexity	O
to	O
the	O
interpretation	O
of	O
EEG	B-Application
signals	O
,	O
rendering	O
linear	O
methods	O
(	O
methods	O
mentioned	O
above	O
)	O
limited	O
.	O
</s>
<s>
Since	O
1985	O
when	O
two	O
pioneers	O
in	O
nonlinear	O
EEG	B-Algorithm
analysis	I-Algorithm
,	O
Rapp	O
and	O
Bobloyantz	O
,	O
published	O
their	O
first	O
results	O
,	O
the	O
theory	O
of	O
nonlinear	O
dynamic	O
systems	O
,	O
also	O
called	O
'	O
chaos	O
theory	O
 '	O
,	O
has	O
been	O
broadly	O
applied	O
to	O
the	O
field	O
of	O
EEG	B-Algorithm
analysis	I-Algorithm
.	O
</s>
<s>
To	O
conduct	O
nonlinear	O
EEG	B-Algorithm
analysis	I-Algorithm
,	O
researchers	O
have	O
adopted	O
many	O
useful	O
nonlinear	O
parameters	O
such	O
as	O
Lyapunov	O
Exponent	O
,	O
Correlation	O
Dimension	O
,	O
and	O
entropies	O
like	O
Approximate	B-Language
Entropy	I-Language
and	O
Sample	B-Language
Entropy	I-Language
.	O
</s>
<s>
The	O
implementation	O
of	O
Artificial	B-Architecture
Neural	I-Architecture
Networks	I-Architecture
(	O
ANN	O
)	O
is	O
presented	O
for	O
classification	O
of	O
electroencephalogram	B-Application
(	O
EEG	B-Application
)	O
signals	O
.	O
</s>
<s>
In	O
most	O
cases	O
,	O
EEG	B-Application
data	O
involves	O
a	O
preprocess	O
of	O
wavelet	B-Algorithm
transform	I-Algorithm
before	O
putting	O
into	O
the	O
neural	B-Architecture
networks	I-Architecture
.	O
</s>
<s>
RNN	O
(	O
recurrent	B-Algorithm
neural	I-Algorithm
networks	I-Algorithm
)	O
was	O
once	O
considerably	O
applied	O
in	O
studies	O
of	O
ANN	O
implementations	O
in	O
EEG	B-Algorithm
analysis	I-Algorithm
.	O
</s>
<s>
Until	O
the	O
boom	O
of	O
deep	B-Algorithm
learning	I-Algorithm
and	O
CNN	B-Architecture
(	O
Convolutional	B-Architecture
Neural	I-Architecture
Networks	I-Architecture
)	O
,	O
CNN	B-Architecture
method	O
becomes	O
a	O
new	O
favorite	O
in	O
recent	O
studies	O
of	O
EEG	B-Algorithm
analysis	I-Algorithm
employing	O
deep	B-Algorithm
learning	I-Algorithm
.	O
</s>
<s>
With	O
cropped	O
training	O
for	O
the	O
deep	O
CNN	B-Architecture
to	O
reach	O
competitive	O
accuracies	O
on	O
the	O
dataset	O
,	O
deep	O
CNN	B-Architecture
has	O
presented	O
a	O
superior	O
decoding	O
performance	O
.	O
</s>
<s>
Moreover	O
,	O
the	O
big	O
EEG	B-Application
data	O
,	O
as	O
the	O
input	O
of	O
ANN	O
,	O
calls	O
for	O
the	O
need	O
for	O
safe	O
storage	O
and	O
high	O
computational	O
resources	O
for	O
real-time	O
processing	O
.	O
</s>
<s>
To	O
address	O
these	O
challenges	O
,	O
a	O
cloud-based	O
deep	B-Algorithm
learning	I-Algorithm
has	O
been	O
proposed	O
and	O
presented	O
for	O
real-time	O
analysis	O
of	O
big	O
EEG	B-Application
data	O
.	O
</s>
<s>
EEG	B-Algorithm
analysis	I-Algorithm
is	O
widely	O
used	O
in	O
brain-disease	O
diagnosis	O
and	O
assessment	O
.	O
</s>
<s>
In	O
the	O
domain	O
of	O
epileptic	O
seizures	O
,	O
the	O
detection	O
of	O
epileptiform	O
discharges	O
in	O
the	O
EEG	B-Application
is	O
an	O
important	O
component	O
in	O
the	O
diagnosis	O
of	O
epilepsy	O
.	O
</s>
<s>
Careful	O
analyses	O
of	O
the	O
EEG	B-Application
records	O
can	O
provide	O
valuable	O
insight	O
and	O
improved	O
understanding	O
of	O
the	O
mechanisms	O
causing	O
epileptic	O
disorders	O
.	O
</s>
<s>
Besides	O
,	O
EEG	B-Algorithm
analysis	I-Algorithm
also	O
helps	O
much	O
with	O
the	O
detection	O
of	O
Alzheimer	O
's	O
disease	O
,	O
tremor	B-Application
,	O
etc	O
.	O
</s>
<s>
EEG	B-Application
recordings	O
during	O
right	O
and	O
left	O
motor	O
imagery	O
allow	O
one	O
to	O
establish	O
a	O
new	O
communication	O
channel	O
.	O
</s>
<s>
Based	O
on	O
real-time	O
EEG	B-Algorithm
analysis	I-Algorithm
with	O
subject-specific	O
spatial	O
patterns	O
,	O
a	O
brain	B-Application
–	I-Application
computer	I-Application
interface	I-Application
(	O
BCI	O
)	O
can	O
be	O
used	O
to	O
develop	O
a	O
simple	O
binary	O
response	O
for	O
the	O
control	O
of	O
a	O
device	O
.	O
</s>
<s>
Such	O
an	O
EEG-based	O
BCI	O
can	O
help	O
,	O
e.g.	O
,	O
patients	O
with	O
amyotrophic	O
lateral	O
sclerosis	O
,	O
with	O
some	O
daily	O
activities	O
.	O
</s>
<s>
Brainstorm	O
is	O
a	O
collaborative	O
,	O
open-source	O
application	O
dedicated	O
to	O
the	O
analysis	O
of	O
brain	O
recordings	O
including	O
MEG	B-Algorithm
,	O
EEG	B-Application
,	O
fNIRS	B-Algorithm
,	O
ECoG	O
,	O
depth	O
electrodes	O
and	O
animal	O
invasive	O
neurophysiology	O
.	O
</s>
<s>
The	O
objective	O
of	O
Brainstorm	O
is	O
to	O
share	O
a	O
comprehensive	O
set	O
of	O
user-friendly	O
tools	O
with	O
the	O
scientific	O
community	O
using	O
MEG/EEG	O
as	O
an	O
experimental	O
technique	O
.	O
</s>
<s>
Combined	O
with	O
facial	O
expressions	O
analysis	O
,	O
EEG	B-Algorithm
analysis	I-Algorithm
offers	O
the	O
function	O
of	O
continuous	O
emotion	O
detection	O
,	O
which	O
can	O
be	O
used	O
to	O
find	O
the	O
emotional	O
traces	O
of	O
videos	O
.	O
</s>
<s>
Some	O
other	O
applications	O
include	O
EEG-based	O
brain	O
mapping	O
,	O
personalized	O
EEG-based	O
encryptor	O
,	O
EEG-Based	O
image	O
annotation	O
system	O
,	O
etc	O
.	O
</s>
