Indonesi
an
Journa
l
of El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
1
3
,
No.
1
,
Jan
uar
y
201
9
,
pp.
1
8
6
~
1
90
IS
S
N: 25
02
-
4752, DO
I: 10
.11
591/ijeecs
.v1
3
.i
1
.pp
1
8
6
-
190
186
Journ
al h
om
e
page
:
http:
//
ia
es
core.c
om/j
ourn
als/i
ndex.
ph
p/ij
eecs
Key fr
am
e extr
action usi
ng hybri
d algo
rith
m of
dy
namic
sign l
anguage
Hussein
Ali A
ldel
fy
1
, M
ah
m
ood H
amz
a Al
-
M
uf
r
aj
i
2
, T
h
amir
R.
Sa
ee
d
3
1,2,3
Depa
rtment
o
f
Elec
tr
ical Engi
nee
ring
,
Univ
ersity
of
T
ec
hnolog
y
,
Ir
aq
1
Facul
t
y
of Engi
nee
ring
,
Univ
ersity
of
Mus
ta
nsiri
y
ah
,
B
aghda
d
,
Ir
aq
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
J
un
10
, 2
018
Re
vised N
ov
19
, 2
018
Accepte
d Nov
30
, 201
8
Video
sum
m
ari
za
t
ion
used
in
m
an
y
applicat
ions
such
as
vide
o
object
rec
ogni
ti
on
and
cl
assifi
cation.
Diffe
ren
t
m
et
ho
ds
used
for
this
purpose:
ke
y
fr
ame
ext
r
a
ct
ion
using
edg
e
det
e
ction
and
ke
y
fr
ame
ext
r
a
ct
ion
using
discre
t
e
wave
let
tra
nsform
(D
WT
).
In
vide
o
proc
essing,
num
ero
us
fra
m
es
cont
a
ini
ng
sim
ilar
informati
on
,
thi
s
le
ads
to
tim
e
consum
pti
on
and
slow
proc
essing
spee
d
and
complexi
t
y
.
New
m
et
hod
Hy
br
id
DW
T
-
Edge
Ke
y
Fram
e
Dete
ctio
n
(HD
EKFD
)
o
f
d
y
n
amic
sign
la
nguag
e
pre
se
nte
d
in
th
is
pape
r.
Thi
s
m
ethod
comb
ine
d
t
wo
al
gorit
hm
s
edge
detec
t
ion
a
nd
DW
T
to
inc
re
ase
proc
es
sing
spee
d
b
y
red
uce
th
e
num
ber
s
of
red
undant
fra
m
es.
Sim
ula
ti
on
resul
ts
il
lustr
at
e
the
eff
icient
of
th
e
proposed
al
gori
t
hm
compare
to
oth
er
m
et
hods
.
Ke
yw
or
ds:
D
WT
Ed
ge dete
ct
ion
Hybr
i
d
Key
f
ram
e
Vide
o
Copyright
©
201
9
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed.
Corres
pond
in
g
Aut
h
or
:
Hu
s
sei
n Ali
A
l
delfy
,
Dep
a
rtm
ent o
f El
ect
rical
En
gi
neer
i
ng,
Un
i
ver
sit
y o
f Te
ch
no
l
og
y,
Ir
aq.
Em
a
il
:
hu
ssien
al
du
lfy
@
gm
ai
l
.co
m
1.
INTROD
U
CTION
To
extract
vali
d
inf
or
m
at
ion
from
vid
eo,
wi
thout
any
loss
of
in
form
at
ion
,
m
uch
at
te
ntion
is
bein
g
paid
to
vid
e
o
processin
g
te
chnolo
gy
.
F
or
this
Key
Fr
a
m
e
is
ver
y
us
efu
l
te
ch
nique
.
Extracti
ng
a
s
m
al
l
nu
m
ber
s
of
ke
y
fr
am
es
that
can
a
bs
tract
t
he
co
ntent
of
vi
deo
is
ver
y
im
portant
for
e
ff
i
ci
ent
bro
wsing
an
d
retrieval
in
vide
o
databas
e.
V
ideo
proce
ssin
g
te
chn
i
qu
es
c
an
be
us
e
d
to
r
e
m
ov
e
redu
ndant
fr
am
es
fr
om
vid
eo
so
that
it
will
so
lve
the
prob
l
e
m
of
the
data
storag
e
[1
]
.
K
ey
fr
am
e
extracti
ng
base
d
on
threshold
te
ch
niqu
e
wh
e
re
the
cal
culat
ed
thres
hold
is
br
ou
gh
t
out
in
com
par
ison
with
the
dif
fer
e
nce
histo
gra
m
of
the
i
m
a
ge
[2
]
.
An
e
ff
ect
i
ve
a
nd
e
ff
ic
ie
nt
co
ntent
ba
sed
ke
y
fr
am
e
detection
m
et
ho
d
ba
sed
on
c
onte
nt
analy
sis
an
d
f
eat
ur
e
detect
ion
f
r
om
key
fr
am
e
pr
opos
e
d
in
[
3].
Sele
ct
and
ex
tract
key
fr
am
es
from
new
s
vid
e
o
shot
bas
ed
on
dissim
il
arit
y
between
each
pa
ir
of
c
onsec
utive
f
ram
es
was
prese
nted
i
n
[
4].
Anothe
r
a
ppr
oac
h
for
key
fram
e
s
extracti
on
f
rom
vid
eo
base
d
on
Fa
ber
-
Sc
haude
r
coe
ff
ic
ie
nts
an
d
si
ng
ular
value
dec
om
po
sit
ion
presented
in
[5
]
thr
ough
the
key
fr
am
e
sequ
e
nce
de
te
ct
ion
and
id
entifi
cat
ion
,
th
e
sign
la
ngua
ge
can
be
ra
pi
dly
recog
nized.
At
the
sa
m
e
tim
e
,
sign
la
ngua
ge
is
a
way
fo
r
deaf
people
to
com
m
un
ic
at
e
and
e
xch
a
nge
ideas
thr
ough
fi
ng
e
r
al
ph
a
bet
a
nd
gestu
res
i
ns
t
ead
of
la
ngua
ge
[
6].
Key
f
ram
e
extracti
on
m
et
ho
d
bas
ed
on
un
s
uper
vise
d
c
lusterin
g
a
nd
m
utu
al
com
pa
ri
son
wer
e
de
ve
lop
e
d
i
n
[
7].
A
m
et
ho
d
of
key
f
ram
e
extracti
on
us
in
g
thres
holdin
g
of
a
bs
ol
ut
e
diff
ere
nce
of
histogram
of
c
on
s
ecuti
ve
f
ra
m
es
o
f
vid
e
o
da
ta
was
pr
op
ose
d
[8
]
.
Key f
ram
e extracti
on
m
et
ho
d usin
g D
WT wa
velet
stat
ist
ic
s w
as
pro
po
se
d
i
n
[
9]. T
he key
fr
am
es are found b
y
com
pu
ti
ng
t
he
edg
e
dif
fe
ren
c
e
betwee
n
t
he
consecuti
ve
f
ra
m
es
and
t
hose
fr
am
es
exceedi
ng
the
t
hr
es
hold
are
cons
idere
d
a
s
key f
ram
e [1
0]
.
In
t
his
stu
dy,
a
Ne
w
Me
th
od
Hybr
i
d
D
WT
-
Ed
ge
Key
Fr
a
m
e
Detect
ion
(
HD
E
KFD
)
of
dynam
ic
sig
n
la
nguag
e
is
pro
po
s
ed
.
This
m
e
thod
com
bin
ed
two
al
gorithm
s
edg
e
detect
io
n
an
d
D
WT
to
increase
pr
oce
ssin
g
sp
ee
d
by
reduc
e
the
num
ber
s
of
redu
nd
ant
fra
m
es.
Si
m
ulatio
n
res
ults
il
lustrate
the
ef
fici
ent
of
the
pro
pose
d
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Ke
y fram
e
extr
ac
ti
on usi
ng
hy
br
id
a
l
go
rit
hm
of
dy
namic
sig
n
la
ngua
ge
(
H
us
sei
n Al
i Al
de
lf
y
)
187
al
gorithm
co
m
par
e
t
o
oth
er
m
et
ho
ds.
T
his
pap
e
r
is
orga
nized
as
fo
ll
ows
.
Sect
ion
2
s
ho
ws
s
om
e
m
e
tho
d
f
or
key
f
ram
e
extracti
on
.
Sect
ion
3
prese
nts
the
pro
posed
m
e
thod,
a
nd
Sect
io
n
4
bri
e
fly
disc
us
se
s
the
exp
e
rim
ental
resu
lt
s
an
d
S
ect
ion
5
is
the
c
oncl
us
i
on
of
t
hi
s
stud
y
in
c
ompare
with
the
D
WT
wa
velet
m
et
ho
d
and the
Ed
ge d
et
ect
ion
m
et
ho
d.
2.
GENER
AL
M
ET
HOD
S
FO
R
KE
Y
F
R
A
ME E
X
T
RAC
TION
This secti
on
de
scribe
d
s
om
e o
f gene
ral m
et
ho
ds u
sed
for ke
y fr
am
e extracti
on
:
2
.
1.
Ke
y Fra
me Ex
tra
c
tion B
as
ed
on
DW
T Wa
velet
Sta
tistics:
The
first
ste
p,
t
he
vid
e
o
fr
am
es
are
rea
d
a
nd,
then
t
wo
c
on
se
cutive
fr
am
es
trans
f
or
m
ed
with
D
WT
t
o
ob
ta
in
f
our
sub
b
a
nds,
LL
,
HL,
L
H,
a
nd
HH.
O
nly
thre
e
su
b
-
bands,
H
L,
LH
,
an
d
H
H
a
re
us
e
d
to
detect
key
fr
am
e
.
Fo
r
eac
h
sub
ba
nd
dif
fer
e
nt
value
is
est
i
m
at
ed
by
s
ub
t
racti
ng
detai
l
co
m
po
ne
nt
va
lues
of
c
urr
e
nt
and
nex
t
fr
am
e as s
how
n
in
E
qu
at
i
on 1
.
(
)
=
∑
∑
(
+
1
(
,
)
−
(
,
)
)
=
1
=
1
=
1
,
2
,
3
(1)
wh
e
re
i
an
d
j
a
re
nu
m
ber
s
of
row
a
nd
c
olu
m
n,
are
t
he
HL
,
LH
a
nd
HH
ba
nd
of
gray
im
ages
and
+
1
are
the
HL,
L
H
an
d
HH
ba
nd
of
gray
i
m
ages
+
1
.
I
n
the
sec
ond
s
te
p,
m
ean
and
sta
ndar
d
dev
ia
ti
on
are c
al
culat
ed
f
r
om
ste
p on
e
u
si
ng
diff
e
re
nce
valu
es of s
ub ban
ds
E
quat
io
n 2 a
nd
E
quat
ion 3
.
=
∑
(
(
)
)
−
1
=
1
−
1
(2)
=
[
∑
(
(
)
−
)
2
−
1
=
1
−
1
]
(3)
The
thir
d
ste
p,
th
res
ho
l
d
val
ue
fo
r
each
s
ub
ba
nd
is
com
pu
te
d
by
E
qu
at
ion
4
.
th
e
la
st
ste
p
a
com
par
ison
be
tween
th
res
ho
l
d
an
d
di
ff
e
rence
values
the
n
if
two
di
ff
e
rence
values
of
a
ny
two
s
ub
ba
nd
s
a
re
gr
eat
er
tha
n
ea
ch r
el
at
ed
th
res
ho
l
d
,
t
he
la
st
fra
m
e con
si
der
e
d
as
a
key f
ram
e.
=
+
.
(4)
2
.
2
.
Ed
ge De
tection
In
this
m
et
ho
d
the
vid
e
o
strea
m
is
ta
ken
as
inpu
t
co
nvert
in
to
dif
fer
e
nt
im
ages
.
M
Af
te
r
t
his
i
m
age
is
conve
rted
i
nto
edg
e
detect
e
d
us
in
g
of
one
of
the
ed
ge
dete
ct
ion
m
et
ho
d.
Ca
nn
y
detect
io
n
m
et
ho
d
is
us
ed
in
thi
s
ph
ase
of
key
fr
am
e
ext
racti
on
.
Af
te
r
that
the
edg
e
diff
er
ence
be
tween
tw
o
s
uc
cessi
ve
fr
am
es
are
cal
culat
ed
and
in
this
way
t
he
dif
fer
e
nce
betwee
n
al
l
fr
a
m
es
is
cal
cu
la
te
d
.
The
n
th
e
total
accu
m
ulate
d
diff
e
re
nce
is
ta
ken
a
nd
by
us
in
g
that
the
m
ean
of
t
he
total
vid
e
o
is
cal
culat
ed.
Af
t
er
that
the
st
a
nd
a
r
d
de
viati
on
is
ca
lc
ulate
d
.
T
he
t
hr
es
hold
is
cal
culat
ed
dep
e
ndin
g
on
m
ean
and
sta
ndar
d
de
viati
on
cal
c
ulati
on
then
if
t
he
dif
fer
e
nce
betwe
en
tw
o
s
ucces
sive
f
ram
es
is
gr
eat
e
r
tha
n
the
th
reshol
d
va
lue
the
key
f
ram
e
is
extracte
d.
Fi
gure
1
s
how
t
he
f
lowc
har
t
for
ke
y
fr
am
e
extracti
on
us
i
ng
D
W
T
m
et
ho
d.
Fig
ur
e
2
sho
ws
t
he
flo
w
char
t
of the
ke
y fr
am
e extracti
on
base
d o
n
e
dg
e
d
et
ect
io
n m
et
ho
d.
3.
PROP
OSE
D MET
HO
D
H
YBRID
DWT
-
EDGE
KEY
FRAME
DET
ECTION
The
propose
d
m
et
ho
d
Hyb
rid
D
WT
-
E
dg
e
Key
Fr
am
e
Detect
ion
(
HD
E
KFD)
c
om
pr
ise
s
two
ph
a
se
,
the
fir
st
phase
is
the
key
f
ram
e
extracti
on
ba
sed
on
discrete
wa
velet
m
et
h
od.
T
he
outp
ut
file
s
f
r
om
this
phas
e
is
ta
ken
as
input
to
the
second
phase
that
ke
y
fr
am
e
extracti
on
base
d
on
edg
e
detect
io
n
as
show
n
in
fig
ur
e
.
Figure
3
s
how
the the
bloc
k d
ia
gr
am
o
f key
fram
e extracti
on
base
d on (
H
DEKFD
)
m
et
ho
d.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
1
3
, N
o.
1
,
Ja
nu
a
ry
201
9
:
1
8
6
–
1
9
0
188
Figure
1. Flo
w
char
t
for key
fr
a
m
e extracti
on u
si
ng DWT m
et
hod.
(Phase
1)
Figure
2. Flo
w
char
t
for key
fr
a
m
e extracti
on b
ase
d o
n
e
dg
e
d
et
ect
io
n
m
eth
od
(
ph
ase
2)
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Ke
y fram
e
extr
ac
ti
on usi
ng
hy
br
id
a
l
go
rit
hm
of
dy
namic
sig
n
la
ngua
ge
(
H
us
sei
n Al
i Al
de
lf
y
)
189
Figure
3.
The
blo
c
k diag
ram
of k
ey
fr
am
e extracti
on
base
d o
n (HDE
KFD
)
m
et
ho
d
4.
THE
E
X
PE
RI
MENT
AL R
E
SU
LT
S
In
this
sect
io
n
of
r
esearc
h
is
ind
ic
at
ive
words
belo
ng
t
o
De
af
an
d
dum
b
peo
ple,
t
hese
w
ords
we
r
e
colle
ct
ed
in
A
rab
ic
la
ngua
ge
an
d
I
ra
qi
dial
ect
and
in
div
i
du
al
w
o
r
ds
i
n
colla
borati
on
with
I
ra
qi
Mi
ni
stry
of
la
bour
a
nd
S
oc
ia
l
Af
fairs
.
Mo
re
tha
n
f
or
ty
w
ords
sig
nalle
d
ta
sk
in
de
al
ing
on
e
w
ord
is
use
d
as
a
n
exa
m
ple,
water
word
IS
us
e
d
as
in
pu
t
s
to
the
propo
sed
syst
em
al
s
o
bee
n
us
in
g
MATLAB
pro
gr
am
2013b
i
n
the
i
m
ple
m
ent
at
io
n
of
t
he
propo
sed
al
go
rithm
.
The
water
vide
o
is
11.9
MH
z
an
d
5
sec
on
d
ti
m
e
le
ng
th
and
30
fr
am
e
/se
c
.
Th
e
com
par
at
ive
resu
lt
s o
f
key
f
ram
e
extracti
on
ba
sed
on
(
H
DEKFD
)
m
et
ho
d
with o
the
r
m
et
ho
ds
is i
ll
us
trat
ed In
T
able
1
.
Table
1.
C
om
par
at
ive
Re
s
ults
Metho
d
No
o
f
f
ra
m
es in
vi
d
eo
Key
f
ra
m
e
No
Discrete
Wav
elet
m
e
th
o
d
174
34
Edg
e detectio
n
m
e
th
o
d
174
27
(HDEKF
D
)
m
eth
o
d
174
10
The
key
f
ram
e
extracti
on
of
D
WT
wa
vel
et
m
et
ho
d
sho
wn
in
F
ig
ur
e
4
.
T
he
num
ber
of
im
ages
about
34
im
age.
Fig
ure
5
s
ho
ws
t
he
key
f
ra
m
e
extracti
on
base
d
on
ed
ge
detect
ion
m
et
h
od
an
d
t
he
nu
m
ber
of
i
m
ages
resu
lt
e
d
ab
out
27
im
a
ges
.
Fi
gure
6
s
hows
the k
ey
f
ram
e
extracti
on
base
d
on
the (
H
DE
KFD)
propose
d
m
et
ho
d
a
nd t
he
num
ber
of im
ages
res
ulted a
bout
(10) im
ag
es
.
Figure
4.
Key
f
ram
e
extracti
on
base
d on disc
rete wa
velet
m
et
hod
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
1
3
, N
o.
1
,
Ja
nu
a
ry
201
9
:
1
8
6
–
1
9
0
190
Figure
5.
Key
f
ram
e extracti
on
base
d on Ed
ge
d
et
ect
io
n
m
eth
od
Figure
6. Key
f
ram
e extracti
on
base
d on H
D
EKFD m
et
ho
d
5.
CONCL
US
I
O
N
In
this
pap
e
r
key
fr
am
e
extracti
on
base
d
on
(HDE
KFD
)
m
e
tho
d
is
presente
d
.
T
he
Aim
of
this
m
et
ho
d
is
to
reduce
the
r
edun
dan
t
f
ra
m
es
that
can
le
ad
dim
ensi
on
al
it
y
reducti
on
of
feat
ur
e
vector
cl
assifi
cat
ion
of
an
isolat
ed
A
rab
ic
w
ord
for
deaf
a
nd
du
m
p
pe
ople
s.
Acc
ord
in
g
to
the
e
xp
e
rim
ental
re
su
lt
s
,
the
(HDE
KFD
)
Me
thod
is
ef
fecti
ve
in
re
duct
ion
of
key
fra
m
es
in
co
m
par
iso
n
with
tw
o
m
e
tho
ds
an
d
this
resu
lt
le
a
ds
t
o
an
e
ff
ect
ive
re
du
ct
io
n
in
ti
m
e
co
nsum
ing
in
processi
ng
tim
e
an
d
re
du
ct
io
n
of
sto
rage
ca
pacit
y
.
As
our
f
uture
work,
we
will
con
ti
nue
in
ou
r
researc
h
f
or
patte
rn
rec
ogni
ti
on
o
f
sign
la
ngua
ge
base
d
on
th
e
resu
lt
ed
k
ey
fra
m
es f
r
om
a u
sed vide
o
cl
ips
.
REFERE
NCE
S
[1]
Am
it
kum
ar
M.
Satput
e
,
K.
R.
Khada
rka
r
”
Video
Summ
ari
za
ti
on
b
y
Removing
Duplicate
Fram
es
from
Surveil
la
nc
e
Video
Us
ing
Key
fr
ame
Ext
r
ac
t
io
n”
Inte
rn
ational
Journal
of
Innov
at
iv
e
Rese
arc
h
i
n
computer
and
Com
m
unic
at
io
n
Engi
ne
eri
ng
,
Vo
l
.
5,
pp
.
8501
-
850
9,
Iss
ue
4
,
April
2017.
[2]
Sanjo
y
Ghat
ak,
“
Ke
y
–
Fram
e
Ext
racti
on
Us
ing
Thre
sholdt
ec
hni
que
”
,
Int
ern
a
ti
on
al
journal
of
Eng
ine
er
ing
Applie
d
Scie
nc
es
and
Technol
og
y
,
Vol.
1
,
Iss
ue
8,
ISS
N No.2455
-
2143,
pp
.
51
-
56,
2016.
[3]
Zhe
n
y
u
W
u,
Rui
quing
W
u
,
Hon
gy
ang
Yu,
Bin
Tang
“
Ke
y
Fram
e
Ext
ra
ct
ion
Towa
rd
s
Kerne
l
–
SIF
T
Ide
nt
ifi
c
ation
”
Inte
rna
ti
ona
l
Confer
ence
o
n
Advanc
ed
Com
pute
r
Scie
nce
and
El
e
ct
ron
i
cs
Infor
m
at
ion,
pp.
163
-
166
,
(IC
ACS
EI
2013).
[4]
Sanjo
y
Ghata
k
,
Debotosh
Bhat
t
a
cha
rj
ee
“
Extrac
tion
of
Ke
y
Fram
es
from
News
V
ide
o
Us
ing
EDF
,
MD
F
AN
D
HI
Method
for
News
Video
Summ
ari
zation
“
Inte
rn
a
ti
onal
Journ
al
of
Engi
ne
eri
ng
an
d
Innova
ti
v
e
Technol
og
y
(IJEIT
)
,
Vol
2,
pp
.
188
-
19
4
,
Iss
ue
12
,
June
2013.
[5]
As
s
m
a
Aze
roua
l
,
Kari
m
Afdel
,
El
Haj
ji
Moham
ed
Hass
an
Douz
i
“
Ke
y
Fram
e
Ext
ra
ct
ion
using
Faber
–
Sch
aud
er
W
ave
le
t
”
Advan
ce
s in
informat
io
n
s
ci
en
ce a
nd
co
m
pute
r
Engi
n
ee
r
ing
.
pp
.
317
-
322,
2015.
[6]
L.
Shurong,
H.
Yuan
y
u
an,
H
.
Z
uoji
n,
an
d
D
.
Q
un,
"K
e
y
Fram
e
Detect
ion
Algo
rit
hm
base
d
on
D
y
namic
Sign
La
nguag
e
Video
for
the
Non
Sp
ec
if
ic
Popul
at
io
n,
"
Inte
rn
at
ion
al
Journal
of
Sign
al
Proce
ss
ing
,
I
m
age
Proce
ss
ing
and
Pat
te
rn
Recogniti
on,
vol
.
8
,
pp.
135
-
148
,
20
15.
[7]
N.
J.
Janwe
an
d
K.
K.
Bho
y
a
r,
"V
ide
o
Ke
y
-
Fram
e
Ext
racti
o
n
using
Uns
up
erv
ised
Clust
ering
and
Mutual
Com
par
ison,"
In
te
rna
ti
ona
l
Journ
al
of
Im
age Proc
essing
(IJIP
),
vol
.
10
,
p
.
73
-
84
,
2
016.
[8]
C.
Sheen
a
and
N
.
Nar
a
y
ana
n
,
"K
e
y
-
fra
m
e
Ext
ra
ction
b
y
Ana
l
y
s
is
of
Histograms
of
Video
Fram
es
Us
ing
S
ta
ti
sti
cal
Methods,
" Procedia
Com
pute
r
Sc
ie
nc
e, vol
.
70,
p
p.
36
-
40
,
2015
.
[9]
K.
T.
Ti
n
t
and
K.
Soe,
"K
e
y
fr
ame
ext
ra
ction
for
vide
o
sum
m
ari
za
ti
on
using
DW
T
wave
let
statist
ic
s,"
Inte
rn
at
ion
al
Journal
of
Advanc
ed
Res
earch
in
Com
pute
r
Engi
n
ee
r
i
ng
&
Te
chno
log
y
(I
JA
RCE
T),
vol
.
2,
p
p.
pp:
1829
-
1833,
2013.
[10]
K.
Khurana
and
M.
Chand
ak,
"K
e
y
fra
m
e
ext
ra
ct
ion
m
e
thodol
o
g
y
for
vid
eo
an
nota
ti
on
,
"
Int
ern
at
ion
al
Journa
l
of
Com
pute
r
Engi
n
ee
ring
and
Tech
nolog
y
,
vo
l. 4, p
p.
221
-
228
,
201
3.
Evaluation Warning : The document was created with Spire.PDF for Python.