TELKOM
NIKA
, Vol. 11, No. 9, September 20
13, pp.
5409
~54
1
4
ISSN: 2302-4
046
5409
Re
cei
v
ed Fe
brua
ry 22, 20
13; Re
vised
June 14, 20
13;
Accept
ed Ju
ne 24, 201
3
Vigilance Degree Computing based on
EEG
Zhen
z
hong Zhan*
1
, Zhen
dong Mu
2
1
Institute of Jiang
xi U
n
ivers
i
t
y
of
T
e
chnol
og
y T
r
aining
Ce
nte
r
Jingd
ong, N
a
ncha
ng, Jia
n
g
x
i 330
02
9
2
Institute of Informatio
n
T
e
chnol
og
y, Jia
n
g
x
i
Universi
t
y
of T
e
chn
o
lo
g
y
, Na
ncha
ng, Jia
n
g
x
i 330
02
9, Chin
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: 4186
23
577
@
qq.com
A
b
st
r
a
ct
In dai
ly life, l
o
ts of w
o
rk need
peo
ple
maint
a
in hi
gh
er attent
ion
or vig
ila
nce
.
In the early s
t
udy o
f
vigil
anc
e, b
link
frequ
ency, th
e i
m
ped
anc
e
of skin,
bo
dy
te
mper
ature
and
bl
oo
d pre
ssure an
d
ot
h
e
r
physi
olo
g
ica
l
si
gna
ls w
a
s use
d
to esti
mate t
he vig
ila
nce. E
E
G signal c
an
mor
e
dir
e
ctly reflect the bra
i
n
'
s
activity
tha
n
other physi
ol
ogic
a
l
s
i
gn
als, an
d EEG
sign
al
ha
ve a
h
i
gh
er ti
me res
o
luti
on. In
this
pa
per, E
R
P
compo
nent a
n
d
differe
nt freq
uenc
ies of EE
G w
e
re used t
o
an
aly
z
e
th
e
alert state, acc
o
rdi
ng to this s
t
udy,
in the E
R
P co
mp
on
ents, N1
70 ca
n b
e
a
goo
d re
prese
n
t
ation of th
e d
egre
e
of fatig
u
e
of the s
ubj
e
c
t;
T
h
roug
h th
e
1
0
su
bjects
EE
G freque
ncy
di
stributio
n a
n
a
l
y
s
is, an
d
accor
d
in
g to
the
for
m
u
l
a
defi
n
e
d
i
n
this
pap
er, the vigi
l
ance d
egr
ee of
this ten subj
ec
ts w
a
s calculated.
Ke
y
w
ords
: ele
c
troence
p
h
a
lo
gra
m
, vigil
anc
e
degre
e
, event-
r
elate
d
pote
n
ti
al, freque
nci
e
s distrib
u
te
Copy
right
©
2013 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
In daily life, lots of work
need
peo
ple
maintain hi
gh
er attention
or vigilan
c
e,
su
ch a
s
long-dista
n
ce
bu
s d
r
ivers, stud
ents'
le
arnin
g
, le
ctures, exami
nati
on. On
ce th
e
y
appea
r i
n
the
work of the
attention
n
o
t ce
ntrali
ze
d, may
cau
s
e
v
e
ry
se
rio
u
s co
ns
equ
e
n
ce
s. T
herefore,
resea
r
ch alert
on man, has
the importa
nt pra
c
tical
signi
fican
c
e.
In the early
stage
s of vigilance, blin
k fre
quen
cy, the impedan
ce of skin, body
temperature
and bl
ood
pressure
a
nd other physi
ol
ogical
si
gnal
s wa
s
u
s
ed
to
estimate
the
vigilance. Erg
onomi
cs
re
se
arch sho
w
s t
hat, when
a person i
s
in a
state of high
alert, and p
a
l
m
ski
n impe
dan
ce will
decre
ase, an
d wh
en peo
ple a
r
e in
a state
of fatigue, skin re
sista
n
ce
will
rise.
In the
study of
Ji
et al
[1], thro
ugh
the a
c
cu
rate
positio
ning
of
the fa
ce,
mo
uth, no
se,
eyes,
eyes
clo
s
ed,
clo
s
ing
time,
blink freq
uen
cy, nod
ding f
r
eque
ncy, fa
ce toward
s, g
a
z
e
dire
ction
a
n
d
mouth o
penin
g
deg
re
e an
d
other fe
ature
s
of vi
gilan
c
e
on h
u
man
re
sea
r
ch. In fact, people
start
to study
exe
m
plary i
n
the
ninetee
n fifties [2
-4], at
first, the stu
d
y st
arts from
sle
e
p proble
m
s,
a
n
d
the
main
diff
eren
ce between wa
king and slee
p
re
se
a
r
ch in two different
co
ndition
s. Wit
h
the
further re
se
arch, p
eople
fro
m
wa
keful
n
e
s
s to sl
eep
is com
p
o
s
ed
of
seve
ral
stag
es, the
analy
s
is
of these
stag
es
whi
c
h feat
ure
s
be
co
me
s the fo
cu
s of
the study; un
til now, with t
he expa
nsi
o
n
of
the scope
of the study, the tradition
al from con
sci
ou
sne
s
s to further
sub
d
ivide
the slee
p st
ate
betwe
en, whi
c
h cl
early put
forwa
r
d the vigilan
c
e.
EEG sig
nal
can mo
re
dire
ctly refle
c
t th
e b
r
ai
n'
s
acti
vity than othe
r p
h
ysiolo
gical si
gnal
s.
EEG sign
al b
e
ca
use of its
non-i
n
vasive,
is e
a
sy to
use in the
stud
y, so the a
p
p
lication
of EEG
sign
al
widely
in b
r
ain
-
co
mputer interf
ace
sy
st
em,
physiol
ogical
dete
c
tion [5
-7], rel
a
ted i
n
the
cordon a
r
e
a
, su
ch a
s
fatigue driving EE
G studie
s
hav
e related to [8
, 9].
ERP can o
b
serve brain a
c
tivity in
the proce
s
s of the wind
ow; it wa
s found that t
he ERP
comp
one
nt is closely relat
ed to many a
nd co
gniti
ve pro
c
e
ss. Fo
r
example: CNV slow p
o
ten
t
ial
comp
one
nts
of British
neu
rophy
siolo
g
ist
Walte
r
re
p
o
rt (co
n
tinge
nt
negative va
ri
ation), i
s
clo
s
ely
related
with
t
he
h
u
ma
n
to look
forwa
r
d to,
pre
parat
io
n, actio
n
time
ori
entation,
pay attention
to
mental activi
ty;
Sutton propo
sed P30
0
is co
m
p
o
s
ition of ERP and attention, recogniti
on,
deci
s
io
n-m
a
ki
ng, memo
ry and othe
r co
gnitive func
ti
on relate
d to
the Kutas a
nd Hillyard first
prop
osed
N400 p
r
omot
ed; the brai
n mechani
sm of huma
n
langu
age
pro
c
e
ssi
ng
and
unde
rsta
ndin
g
. Acco
rdin
g
to the basic characte
ri
st
ic of vigilance, determin
e
which co
gni
tive
function
in
clu
d
ing
attentio
n, memo
ry,
motivati
on, a
nd e
m
otion,
cog
n
itive fun
c
tion
s
su
ch
as
langu
age, co
ntrol, and ERP experiment
al analysi
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 9, September 201
3: 540
9 – 5414
5410
In this pa
per,
ERP com
p
o
nent an
d different frequ
en
cie
s
of EEG
were u
s
ed to
analyze
the ale
r
t
sta
t
e, acco
rding
to this
stu
d
y,
in the
ERP
comp
one
nts, N170
can b
e
a
go
od
rep
r
e
s
entatio
n of the degree of
fatigue
of the subj
ect
;
Throug
h the
10 su
bject
s
EEG freque
n
cy
distrib
u
tion a
nalysi
s
, and
according to t
he formul
a d
e
fined in this
pape
r, the vigilan
c
e de
gre
e
of
this ten su
bje
c
ts was
cal
c
u
l
ated.
2. EEG Acquisition
The EEG d
a
ta used in
this pape
r i
s
co
me from B
C
I
Labo
rato
ry of Jian
gxi Univ
ersity of
Tech
nolo
g
y; the subj
ect
s
to relax in a quiet shie
ldi
n
g
room sat an
armless chai
r in front of the
comp
uter, wa
tching the
screen, do EEG
experime
n
t according to
experim
ent the arran
gem
en
t
and the in
dication screen
stimulation.
EEG acq
u
is
it
ion is the
use of 40
Neuroscan a
m
plifi
e
r,
were obtai
ne
d by scan
4.3
softwa
r
e,
rig
h
t mast
oi
d is
referen
c
e el
e
c
trod
e, an
d u
s
ed
100
0Hz
as
sampli
ng rate
, band acqui
sition usin
g 20
0Hz low-pa
ss, high-pa
ss 0.05Hz an
d 50
Hz n
o
tch.
P300 was di
scovere
d
by
Sutton in 19
65, its
mai
n
feature
is
a forward wave
event in
about 3
00 mi
llise
c
on
ds, e
ndog
enou
s
compon
ents
main
ly and
p
s
ych
o
logi
cal f
a
ctors
relate
d, its
physi
cal mea
n
ing is mainl
y
reflected in
the latenc
y of the subje
c
ts of stimuli or cla
s
sificati
on
requi
re
d time
, amplitude
said reflecte
d
backg
rou
nd
or m
e
mo
ry u
pdate
s
. So t
he ale
r
t a
nal
ysis
can u
s
e P30
0
;
EEG is comp
ose
d
by vario
u
s b
and
wav
e
s, the frequ
ency
can
be
divided into
,,
,
.
The
,
is slo
w
wave, occu
rs mainly in ad
ults sle
ep; an
d
,
is fast wave, occurs mai
n
ly in
peopl
e is vigilant and pay a
ttention external stim
ulu
s
o
r
whe
n
the sp
ecial me
ntal activity.
In this
pape
r, electroen
ce
phalo
g
ra
m E
E
G
sel
e
cte
d
10 colle
ge stude
nts as subj
ect
s
,
inclu
d
ing
five su
bje
c
ts i
n
t
he m
o
rni
ng t
o
do
the
exp
e
rime
nt, five partici
pant
s d
i
d the
expe
ri
ment
in the afte
rno
on (not a
nap
). The
stimul
u
s
p
a
ttern
: first, the screen i
s
a
se
co
nd
bl
ack
state, the
n
will rand
omly appe
ar in a
pictu
r
e, the
pictur
e
sho
w
s 2
50 milli
se
con
d
s, th
en
a second
bla
c
k
scree
n
, then
will rand
omly
appe
ar i
n
a
p
i
cture,
befo
r
e
the exp
e
rim
e
nt, parti
cipant
s
will b
e
a
s
ke
d
to write test
s in a picture nu
mber.
3. Data Pro
c
essing
This p
ape
r a
nalyze
s
the V
i
gilan
c
e de
gree from the freque
ncy an
d
ERP com
pon
ents, so
data pro
c
e
s
si
ng, respe
c
tively:
The ERP co
mpone
nts an
alysis
step is:
Step 1, Block larg
er d
r
ift EEG: in EEG acq
u
isiti
on pro
c
e
s
s, su
ch a
s
the
subje
c
ts
movement, wander, out
side sound
effects, the EEG signal initia
l
there
will
be large drift, will
follow the EEG sign
al pro
c
essing imp
a
ct, so in t
he EEG before tre
a
tment, to remove this a p
a
rt
of the brain el
ectri
c
al si
gnal
;
Step 2, Ocula
r
artifact
redu
ction: the EEG sign
al in th
e origi
nal, be
cau
s
e to bli
n
k or lo
ok
right and lef
t, the impact on the eye electri
c
si
gnal, so bef
ore featu
r
e extraction a
n
d
cla
ssifi
cation,
to remove
th
e imp
a
ct
of t
h
is
pa
r
t, this
pape
r i
s
mai
n
ly to remov
e
the
vertical
eye
film;
Step 3, epo
ch the data, view the
stimul
us in
te
rvals,
and ge
ne
rally
10%-20%, -5
0, is the
comm
on val
u
e of -100; th
e
spi
r
it of n
o
t
more
than
on
e event to
th
e pri
n
ci
ple, in
this
pap
er, t
h
e
interception o
f
data in -100
~90
0
ms.
Step 4, ba
sel
i
ne
corre
c
tio
n
: the
segm
e
n
ted d
a
ta ma
ny not at
ba
selin
e, so thi
s
p
ape
r
con
d
u
c
ted a total of two times the ba
se
line co
rrectio
n
and a line
a
r corre
c
tion.
Step 5, artifact reje
ct: EEG sign
als
coll
ect
ed by the cla
ss, there
is a part of se
gmente
d
data cau
s
e
d
by various re
aso
n
s is n
o
t good, not onl
y for data analysis u
s
ele
s
s, it will affect th
e
analysi
s
of the data, so to cho
o
se a ce
rtain win
dow
scre
enin
g
, the wind
ow i
s
-80
~
80.
Step 6, average: Thi
s
art
i
cle mainly is
ERP analysis of EEG si
gnal
s, so the
same
stimulation of
the brain ele
c
tri
c
al
sig
nal types were
stacked.
The freq
uen
cy analysis
ste
p
is:
Step 1: Subj
ects
duri
ng t
he test p
r
o
c
e
ss, e
s
p
e
ci
ally in fatigue, u
n
co
nscio
u
sly
gene
rate
the moveme
nt of the bod
y, resulting E
E
G interfer
en
ce by EMG,
so the first st
ep is
sele
ct the
better EEG.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Vigilance De
gree
Com
puti
ng ba
sed o
n
EEG (Zhen
zh
ong Zha
n
)
5411
Step 2: On EEG analyzes,
the prima
r
y band di
stri
but
ion 1~50
Hz i
n
this pa
per,
so befo
r
e
do data an
alysis mu
st filter the EEG sign
al.
Step 3: Com
m
on ave
r
ag
e
:
In this p
ape
r, we
u
s
e
Hjo
r
t de
rivation t
o
re
du
ce inte
rfere
n
ce
from the neig
hbori
ng ele
c
trode,
The Hjo
r
t deri
v
ation
H
i
C
is cal
c
ulated a
s
:
1
(1)
4
i
H
ii
j
jS
Cc
s
c
Whe
r
e
ci is t
he re
ading
o
f
the cente
r
electrode
scj, with i=1
…
3
0
and j is th
e set of
indices
corre
s
po
ndin
g
to the eight ele
c
t
r
ode
s surrou
nding el
ectro
de ci.
Step 4: AR
conve
r
si
on: time-d
omain
EEG
data di
sorgani
ze
d E
E
G in orde
r
to better
highlight th
e
cha
r
a
c
teri
stics of EEG
sig
nal, we
u
s
e A
R
mo
del to
convert the tim
e
dom
ain
sig
nals
into freque
ncy domain, an
d extract the feat
ure from the frequ
en
cy domain
sign
a
l
s.
4. Results a
nd Discu
ssi
on Conclu
sion
As sho
w
n in
Figure 1, the uppe
r left corner
Subje
c
t said gra
phi
cs
para
m
eters; the uppe
r
right
co
rne
r
i
s
the a
c
qui
siti
on time,
HEO
is
hori
z
o
n
tal
EOG, VEO i
s
vertical
EOG
,
electrode
F
P
1
labele
d
10-2
0
unde
r different. The gre
en line is
the mornin
g su
bject
s
wa
king
state made the
experim
ent, red line is
su
bject
s
in the ca
se did n
o
t take a na
p, have been q
u
i
e
t made
to wait
until four p.
m. experimental EEG. FT7 aftern
oon FIG electrode deviation volatility was
signifi
cantly h
i
gher than
th
e re
d, green
corre
s
p
ondin
g
EEG ERP
comp
one
nts
also
obviou
s
from
the other ele
c
trod
es
can
be appa
re
nt, in orde
r to
be
tter able to look at two stat
es contra
st, we
cho
o
se one o
f
the electrod
es to discu
s
s.
Figure 1. Two
States Comp
arison Cha
r
t
As sho
w
n in
Figure 2, g
r
een
for
wa
king st
ate,
re
d for m
ental
fatigue, First N170,
rep
r
e
s
ent
s
a degree of
refl
ection of
the subj
ect
of
th
e
pictu
r
e, a
green
refl
e
c
ted on
the pictu
r
e
it
is clea
r,
the
amplitude
re
ach
ed 7.5
mi
crovolt, while
the red
-
N17
0
only 2.5
mi
crovolt, only one-
third of the g
r
een, the info
rmation p
r
o
c
essi
ng P3
00
blue ma
gnitu
de more obvi
ous, but the
red
P300 is very
obviou
s
su
bje
c
ts ba
si
c loss
of informatio
n fatigue pro
c
essing
cap
a
ci
ty.
Re
spe
c
tively on the five
wa
king
state
subj
e
c
ts a
n
d
five fatigue state subj
ects EEG
analysi
s
, the
five wakin
g
state a
c
qui
sition of EEG
, its N1
70 am
p
litude 7.5,
5.3, 6.5, 7.3, 6.8
microvolt, the
mean
6.68
is microvolt. Fi
ve fatigue
N170
amplitud
e, re
spe
c
tivel
y
2.5, 2.0, 1.
9,
3.2, 2.8 microvolt, the mean is 2.4
8
microv
olt.
Therefore, subje
c
ts EEG
N170 am
plitude
analysi
s
, it
ca
n be
con
s
ide
r
ed
whe
n
the
N17
0
a
m
plitu
de tha
n
5
mi
crovolt, subje
c
ts in
the
wa
ki
ng
state, the
sub
j
ects ha
d a
hi
gher vigilan
c
e de
gree,
whi
l
e N170
ma
g
n
itude
of le
ss than
3 mi
cro
v
olt
whe
n
the su
b
j
ect is fatigue
state, s
ubj
ect
s
’ vigilan
c
e d
egre
e
is lo
we
r.
HE
O
G
VE
O
G
FP
1
FP
2
F7
F3
FZ
F4
F8
FT
7
FC
3
FC
Z
FC
4
FT
8
T3
C3
CZ
C4
T4
TP
7
CP
3
CP
Z
CP
4
TP
8
T5
P3
PZ
P4
T6
O
1
O
Z
O
2
Su
b
j
ect:
E
E
G
f
i
l
e
:
22
.
a
vg
Re
c
o
r
d
e
d
:
1
1
:
3
4:
16
19
-D
e
c
-
2
0
1
2
R
a
te - 1
0
0
0
Hz,
HP
F -
0
Hz,
L
P
F - 3
0
0
Hz,
No
tch
- 5
0
Hz
Neu
r
o
s
ca
n
SC
A
N
4
.
3
P
r
i
n
t
e
d
:
1
7
:
3
2:
55
29
-Ja
n
-2
01
3
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 9, September 201
3: 540
9 – 5414
5412
Figure 2. Single Electrode
Comp
ari
s
o
n
s
The P3
00
co
mpone
nt of t
hese 1
0
subj
ects, f
r
om
th
e time
of o
c
curren
ce
and
magnitud
e
of the five
wa
king
state
su
bject
s
, the
P3
00 a
m
plitude
are
2.4, 2.7,
1.8, 2.1,
1.5
microvolt, time of
occurrence are di
stributed in 300 milli
seconds to
350 milliseconds. Five
fatigue state
subjects’
P300 amplit
ude dist
ributi
on betwe
en
0.5 and 1.5,
there are t
w
o su
bje
c
ts P300 from
th
e
sup
e
rim
p
o
s
e
d
pictu
r
e poi
nt of view no
t out. From the P300 a
p
p
ear time, the
time distrib
u
t
ion
betwe
en 32
0 millise
c
on
ds
and 40
0. So, to vigilance d
egre
e
the N1
70 is better th
an P300.
The m
a
in
co
mpone
nt of t
he b
r
ai
n
wav
e
signal
by
four
ba
nd
s:
,,
,
, the domin
ant
comp
one
nt whe
n
human
conscio
u
sn
ess is
,
; the dominant compon
ent when hum
an
uncon
sci
ou
sn
ess i
s
,
.. The
different EEG
ban
ds corre
s
po
ndin
g
to t
he freque
ncy
ra
nge
are:
8~1
2
Hz, fo
r 1
3
~3
0Hz, 4
~
7
H
z,
1~3Hz. In
this pap
er
, th
e mo
del
2, th
e subje
c
ts’
ey
es
clo
s
e
d
stat
e
colle
cted EE
G analysi
s
.
This
pap
er
u
s
ing
the follo
wing
formul
a
to cal
c
ulate
the p
r
op
ortion
of S differe
n
t
states,
different freq
uen
cie
s
:
2
1
50
1
()
(
)
(2)
()
At
d
t
S
At
d
t
is different freque
ncy ba
n
d
.
1,
2
are the lo
wer
and u
p
p
e
r limits of th
e band.
()
A
t
is EEG sign
al
function after AR model tra
n
sformation.
In this pape
r, the following
formula u
s
ed
to ca
lculate the pro
portio
n
of different band
s of
the five subje
c
ts of the two
states:
5
1
()
(
)
(3)
5
S
S
This a
r
ticle u
s
ing the follo
wing formula
to calculate the su
bje
c
ts al
ert deg
ree
s
AL:
()
(
)
(4)
()
(
)
SS
AL
SS
Interce
p
t the
EEG coll
ecte
d by
subje
c
ts’ clo
s
e th
eir e
y
es time to
2
46
segm
ents of on
e
se
con
d
as th
e perio
d. Afte
r data pro
c
e
s
sing i
s
perfo
rmed acco
rdin
g to the method of the abo
ve-
mentione
d, the frequ
en
cy data obt
ain
e
d
in the su
pe
rpositio
n averaging o
b
taine
d
10
subje
c
ts in
the EEG average frequ
en
cy after conv
ersi
on of
the
AR model d
a
ta, and the
n
according
to
Equation
2, to cal
c
ul
ate e
a
ch f
r
eq
uen
cy segm
ent p
r
opo
rtion. T
h
e differe
nt freque
ncy b
a
n
d
s
prop
ortio
n
of five wakin
g
st
ate sho
w
a
s
Table 1,
the
different freq
uen
cy band
s
prop
ortio
n
of five
fatigue state
sho
w
a
s
Tabl
e 2.
ms
-10
0
.
0
0.
0
10
0.
0
20
0.0
30
0.0
40
0.
0
50
0.
0
60
0.
0
70
0.
0
80
0.
0
90
0.
0
uV
0.
0
-2
.5
-5
.0
-7
.5
-10
.
0
-12
.
5
2.
5
5.
0
7.
5
10
.0
12
.5
*
2
2.
av
g
22
.
a
v
g
Su
b
j
ect:
EE
G
f
i
l
e
:
2
2
.
a
v
g
R
eco
r
d
ed
:
1
1
:
3
4
:
1
6
1
9
-
Dec-
2
0
1
2
R
a
te -
1
0
0
0
Hz,
HP
F -
0
Hz,
LP
F -
3
0
0
Hz
,
No
tch
-
5
0
Hz
Neu
r
o
s
can
SC
A
N
4
.
3
P
r
i
n
t
e
d :
17
:
3
3:
34
29
-
J
a
n
-
2
0
1
3
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Vigilance De
gree
Com
puti
ng ba
sed o
n
EEG (Zhen
zh
ong Zha
n
)
5413
Table 1. Diffe
rent Freque
n
c
y Bands P
r
o
portion
of Five Waki
n
g
State
1subject 0.3785
0.4516
0.1093
0.0606
2subject 0.3223
0.5073
0.1262
0.0442
3subject 0.3286
0.4690
0.1458
0.0566
4subject 0.3052
0.4993
0.1094
0.0861
5subject 0.3302
0.4595
0.1288
0.0815
average
0.33296
0.47734
0.1239
0.0658
Table 2. Diffe
rent Freque
n
c
y Bands P
r
o
portion
of Five Waki
n
g
State
1subject 0.2915
0.4107
0.1985
0.0992
2subject 0.2949
0.4083
0.2229
0.0740
3subject 0.2574
0.4058
0.2290
0.1078
4subject 0.2671
0.3849
0.2404
0.1078
5subject 0.2887
0.3558
0.2420
0.1135
average
0.27992
0.3931
0.22656
0.10046
Table 1 an
d 2 data sh
ow t
hat whe
n
the subj
ect
s
we
re
awa
k
e,
ban
d prop
ortio
n
of the
total band di
stributed b
e
tween 30.5
2
% and 37.8
5
%, an
average of
33.30%; subj
ects in a fatig
ue
st
at
e
band proportio
n
of the total ban
d distri
b
u
tion
of between
25.74% and 29.49%, an
averag
e of 28.99%; when
the subje
c
ts were a
w
a
k
e
,
band prop
ortion of the
total band
distrib
u
ted be
tween 45.1
6
%
and 50.73
%, an averag
e of 49.93%; subje
c
ts in a
fatigue state
band p
r
o
porti
on of the total band
dist
ri
bution
of bet
wee
n
35.58
% and 41.0
7
%
, an avera
ge of
39.31%; wh
e
n
the su
bje
c
ts we
re a
w
a
k
e,
band p
r
opo
rtion of
the total ban
d distri
buted
betwe
en 1
0
.93% and
14
.58%, an av
erag
e of
12.
39%; subj
ect
s
in a fatig
u
e
state
band
prop
ortio
n
of
the total ban
d
distri
bution
o
f
betwee
n
19.
85% and
24.
20%, an ave
r
age of
22.66
%;
whe
n
the su
b
j
ects
we
re a
w
ake,
band proportio
n
of the total band
distrib
u
ted be
tween 4.4
2
%
and 8.6
1
%, an avera
ge of
6.58%; subj
e
c
ts in
a fatigu
e state
band
prop
ortio
n
of
the total ban
d
distrib
u
tion of
betwee
n
7.4
0
% and 11.3
5
%, an avera
ge of 10.04%
.
Table 1 an
d Table 2 data
sho
w
that the
subje
c
t is a
w
a
k
e; the proportio
n
of band
,
highe
r tha
n
the subje
c
t i
s
fatigue; the
s
ubje
c
t is
awa
k
e; the
pro
p
o
r
tion of b
and
,
lower than
the subj
ect is fatigue; Com
p
reh
e
n
s
ive T
able 1 an
d
2
of these 10
subje
c
ts the d
a
ta acco
rding
to
equatio
n (3)
can
be
cal
c
ul
ated Th
e ten
su
bject
s
al
e
r
t deg
ree
s
were: 4.8
858,
4.8685, 3.9
4
07,
4.1151, 3.75
51, 2.3588, 2
.
3685, 1.969
1,1.8725, 1.8
129, acco
rdin
g to
the state
of the subje
c
t,
whe
n
the
sub
j
ects ale
r
t d
e
g
ree
s
l
o
wer than
3, indi
cat
i
ng that th
e
subje
c
ts
have
been
tired
at
this
time shoul
d remind the su
bject
s
.
5. Conclusio
n
When a person i
s
awake, alert will
be hi
gh;
when people in a
tired state,
human
vigilance performan
ce
will
be followe
d
low, but
what to measure, an
d the
detection al
ert
degree
s, ha
s be
en a p
r
oble
m
. In this p
ape
r, the EEG to
analyze
pe
ople'
s vigila
nce
perfo
rman
ce,
respe
c
tively, cal
c
ulate
d
from the
ERP
and the
pro
p
o
rtion of diffe
rent ba
nd
s. T
h
e
analysi
s
sho
w
ed that 10
subje
c
ts, wh
en
N170 am
plitude than 5 mi
crovolt, su
bje
c
ts in the waking
state, the
sub
j
ects ha
d
a hi
gher al
ert, a
n
d
when
the
N170
amplitud
e le
ss tha
n
3
microvolt wh
en
subj
ect
s
in
a
tired
state
lo
wer
subj
ect
s
Cautio
nary;,
whe
n
the
su
bject
s
al
ert
d
egre
e
s b
e
low 3,
indicating tha
t
the subje
c
ts have been tired, subj
ect
s
should remind
.
In this pape
r, the EEG to
cal
c
ulate the
subj
ect
s
alert
,
extends the
degre
e
of vigilan
c
e
resea
r
ch a
p
p
r
oa
ch,
but th
e results
of this
pap
er onl
y 10
subj
ect
s
analy
s
is,
su
bject
s
we
re from
Jian
gxi Institu
t
e of Te
ch
nol
ogy in
the
col
l
ege
st
ud
ents,
the
n
u
mbe
r
of
the study popul
ation an
d
subj
ect
s
ag
e
distrib
u
tion i
s
less a
singl
e
,
ther
efore, to expan
d the
numbe
r of
subje
c
ts a
nd t
h
e
wealth of subj
ects a
ge di
stribution is
the
next step in this re
se
arch
work.
Ackn
o
w
l
e
dg
ements
This
wo
rk wa
s
supp
orte
d
by IT proj
ect
s
of
Jian
gxi
Office of Ed
u
c
ation [
N
o. G
J
J102
73]
and Nature of
Jiangxi University of Tech
nology [No.X
YKJ100
4]
Referen
ces
[1]
Ji Q, Yan
g
X.
Real-tim
e
e
y
e,
gaz
e, an
d fac
e
p
o
se track
i
n
g
for mo
nitor
i
n
g
driv
er vi
gil
a
n
c
e.
Re
al-Ti
m
e
Im
aging
, 200
2, 8(5): 357-3
77.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 9, September 201
3: 540
9 – 5414
5414
[2]
Strijkstra A M,
Beersma D G M, Dra
y
er B, et
al. Subj
ective
slee
pin
e
ss corr
elates n
e
g
a
tive
l
y
w
i
t
h
gl
ob
a
l
alp
ha (8
–12 H
z
) and pos
itive
l
y
w
i
t
h
centra
l frontal
theta (4
–8 Hz) frequ
e
n
cies i
n
the h
u
man resti
n
g
a
w
ak
e electro
e
n
cep
hal
ogr
am.
Neurosc
i
enc
e letters
, 2003, 3
40(1): 17-
20.
[3]
Shi L C, Yu H, Lu B L. Semi-s
upervised clustering for vigila
nce
analy
sis based on EEG[C]//Neural
Net
w
orks, 20
0
7
. IJCNN 200
7
.
Internation
a
l
Joint Co
nfere
n
c
e on.
IEEE, 2007: 15
18-
152
3.
[4]
Yu H, Shi L C, Lu B L. V
i
gilanc
e Estimation
Bas
ed on EEG Signals[C
]//Proceedings of the.
Internatio
na
l C
onfere
n
ce o
n
Mecha
n
ica
l
En
gin
eeri
ng 2
007
(ICME2007). 2
007.
[5]
Ange
lakis
E,
Hatzis A, P
a
n
ouri
a
s I G, et
al. Br
ai
n-c
o
m
puter
interfac
e
:
a rec
i
proc
al
self-reg
u
late
d
neur
omodulati
on[M]//Operative Neur
omod
ulation. Springer
Vienna, 2007: 555-
559.
[6]
Bay
liss J D, Inverso S A,
T
entler A. Chan
gin
g
the P30
0
bra
i
n comp
uter int
e
rface.
CyberP
sychol
ogy
&
Behav
ior
. 20
04
; 7(6): 694-70
4
.
[7]
Bianc
hi L, Qu
itadam
o L, Garreffa G, et
al. Perf
orman
c
es eval
uati
o
n
and o
p
timiza
tion of bra
i
n
computer
interf
ace s
y
stems
in
a co
p
y
s
pel
lin
g task.
N
eura
l
Systems an
d Reh
abi
litatio
n Engi
neer
in
g
,
IEEE Transactions on
. 2
007;
15(2): 20
7-2
1
6
.
[8]
W
U
S, GAO L, W
A
NG L. Det
e
cting
Driv
i
ng F
a
tigue Base
d on
El
ectroe
nce
pha
logr
am.
T
r
ansacti
ons of
Beiji
ng Institute
of T
e
chnolo
g
y
. 2009; 12: 00
8
.
[9]
Lin C T
,
W
u
R C, Lia
ng S
F
,
et al. EEG-based
dro
w
siness
estimat
i
on for s
a
fet
y
drivin
g us
in
g
ind
epe
nd
ent compo
nent a
nal
ysis.
Circ
u
its and Syste
m
s I: Regul
ar Pap
e
rs, IEEE Transactio
n
s on
.
200
5; 52(1
2
): 2726-
273
8.
Evaluation Warning : The document was created with Spire.PDF for Python.