TELK
O
M
N
I
K
A
Te
l
e
c
ommu
n
i
c
at
i
on
,
C
omp
u
ti
n
g
,
El
e
c
t
r
on
i
c
s
an
d
C
on
tr
o
l
V
ol
.
18
,
N
o
.
2
,
A
pri
l
2020
,
p
p.
761
~
769
IS
S
N
:
1693
-
6930
,
a
c
c
r
e
di
t
e
d
F
i
rs
t
G
r
a
de
by
K
e
m
e
n
ri
s
t
e
kdi
k
t
i
,
D
e
c
r
e
e
N
o:
21
/
E
/
K
P
T
/
2
018
D
O
I:
10
.
129
28/
T
E
L
K
O
M
N
IK
A
.
v1
8i
2
.
1474
7
761
Jou
r
n
al
h
o
m
e
pa
ge
:
ht
t
p:
/
/
j
our
nal
.
uad
.
a
c
.
i
d/
i
nd
e
x
.
php
/
T
E
L
K
O
MNIKA
A
r
o
b
u
st
me
t
h
o
d
f
o
r
V
R
-
b
a
s
e
d
h
a
n
d
g
e
st
u
r
e
r
e
c
o
g
n
i
t
i
o
n
u
si
n
g
d
e
n
si
t
y
-
b
a
sed CN
N
Li
l
i
an
a
1
,
Ji
-
H
u
n
C
h
ae
2
,
Joon
-
Jae
L
e
e
3
,
Byu
n
g
-
G
ook
Le
e
4
1,
4
C
o
m
pu
t
e
r
E
ng
i
ne
e
r
i
ng
D
ong
s
e
o
U
n
i
v
e
r
s
i
t
y,
S
out
h
K
o
r
e
a
1
I
nf
o
r
m
a
t
i
c
s
D
e
pa
r
t
m
e
nt
P
e
t
r
a
C
hr
i
s
t
i
a
n
U
ni
ve
r
s
i
t
y
,
I
ndon
e
s
i
a
2
,
3
S
c
hoo
l
o
f
C
o
m
pu
t
e
r
E
ngi
ne
e
r
i
ng
K
e
i
m
yung
U
n
i
ve
r
s
i
t
y,
S
o
ut
h
K
or
e
a
A
r
ti
c
l
e
I
n
fo
A
BS
TR
A
C
T
Ar
t
i
c
l
e
h
i
s
t
or
y
:
Re
c
e
i
v
e
d
J
ul
3
,
201
9
Re
vi
s
e
d
J
a
n
16
,
202
0
A
c
c
e
pt
e
d
F
e
b
11
,
202
0
M
a
n
y
V
R
-
b
a
s
e
d
m
e
di
c
a
l
pu
r
po
s
e
s
a
pp
l
i
c
a
t
i
ons
ha
ve
b
e
e
n
de
ve
l
op
e
d
t
o
h
e
l
p
pa
t
i
e
n
t
s
w
i
t
h
m
ob
i
l
i
t
y
d
e
c
r
e
a
s
e
c
a
us
e
d
b
y
a
c
c
i
de
nt
s
,
d
i
s
e
a
s
e
s
,
o
r
ot
h
e
r
i
n
j
u
r
i
e
s
t
o
d
o
phy
s
i
c
a
l
t
r
e
a
t
m
e
nt
e
f
f
i
c
i
e
n
t
l
y
.
V
R
-
ba
s
e
d
a
pp
l
i
c
a
t
i
ons
w
e
r
e
c
ons
i
de
r
e
d
m
o
r
e
e
f
f
e
c
t
i
v
e
h
e
l
pe
r
f
or
i
nd
i
v
i
du
a
l
ph
ys
i
c
a
l
t
r
e
a
t
m
e
n
t
be
c
a
u
s
e
o
f
t
h
e
i
r
l
ow
-
c
os
t
e
qu
i
p
m
e
n
t
a
nd
f
l
e
x
i
b
i
l
i
t
y
i
n
t
i
m
e
a
nd
s
pa
c
e
,
l
e
s
s
a
s
s
i
s
t
a
nc
e
of
a
ph
ys
i
c
a
l
t
he
r
a
pi
s
t
.
A
c
ha
l
l
e
ng
e
i
n
de
v
e
l
opi
ng
a
V
R
-
ba
s
e
d
phy
s
i
c
a
l
t
r
e
a
t
m
e
n
t
w
a
s
u
nde
r
s
t
a
n
di
ng
t
h
e
b
ody
pa
r
t
m
ov
e
m
e
n
t
a
c
c
ur
a
t
e
l
y
a
n
d
qu
i
c
k
l
y
.
W
e
p
r
op
os
e
d
a
r
ob
us
t
pi
pe
l
i
n
e
t
o
un
de
r
s
t
a
nd
i
n
g
ha
n
d
m
o
t
i
o
n
a
c
c
ur
a
t
e
l
y.
W
e
r
e
t
r
i
e
ve
d
our
d
a
t
a
f
r
o
m
m
ove
m
e
nt
s
e
ns
or
s
s
uc
h
a
s
H
T
C
v
i
ve
a
n
d
l
e
a
p
m
o
t
i
on.
G
i
ve
n
a
s
e
q
ue
n
c
e
pos
i
t
i
on
of
p
a
l
m
,
w
e
r
e
p
r
e
s
e
n
t
ou
r
da
t
a
a
s
b
i
n
a
r
y
2D
i
m
a
g
e
s
of
ge
s
t
u
r
e
s
h
a
pe
.
O
u
r
d
a
t
a
s
e
t
c
ons
i
s
t
e
d
o
f
14
k
i
nd
s
o
f
ha
n
d
ge
s
t
u
r
e
s
r
e
c
o
m
m
e
nde
d
by
a
phy
s
i
ot
h
e
r
a
pi
s
t
.
G
i
ve
n
33
3D
p
oi
n
t
s
t
h
a
t
w
e
r
e
m
a
ppe
d
i
n
t
o
bi
na
r
y
i
m
a
ge
s
a
s
i
np
ut
,
w
e
t
r
a
i
n
e
d
ou
r
pr
opos
e
d
d
e
ns
i
t
y
-
b
a
s
e
d
C
N
N
.
O
ur
C
N
N
m
od
e
l
c
on
c
e
r
ne
d
w
i
t
h
our
i
n
pu
t
c
h
a
r
a
c
t
e
r
i
s
t
i
c
s
,
ha
v
i
ng
m
a
ny
'
b
l
a
nk
b
l
o
c
k
pi
xe
l
s
'
,
'
s
i
ngl
e
-
p
i
x
e
l
t
hi
c
kn
e
s
s
'
s
ha
pe
a
n
d
ge
ne
r
a
t
e
d
a
s
a
b
i
na
r
y
i
m
a
ge
.
P
yr
a
m
i
d
ke
r
n
e
l
s
i
z
e
a
ppl
i
e
d
on
t
he
f
e
a
t
u
r
e
e
xt
r
a
c
t
i
on
p
a
r
t
a
n
d
c
l
a
s
s
i
f
i
c
a
t
i
o
n
l
a
ye
r
us
i
ng
s
of
t
m
a
x
a
s
l
os
s
f
u
nc
t
i
o
n,
ha
ve
gi
ve
n
97
.
7
%
a
c
c
u
r
a
c
y.
Ke
y
w
or
d
s
:
2D
i
m
a
g
e
ge
s
t
ure
re
pre
s
e
nt
a
t
i
on
Bi
na
r
y
i
m
a
ge
l
e
a
rni
ng
D
e
ns
i
t
y
-
ba
s
e
d
CN
N
H
a
nd
ge
s
t
u
re
re
c
ogn
i
t
i
on
VR
-
ba
s
e
d
phys
i
c
a
l
t
r
e
a
t
m
e
nt
T
hi
s
i
s
an
op
e
n
ac
c
e
s
s
ar
t
i
c
l
e
u
nde
r
t
he
C
C
B
Y
-
SA
l
i
c
e
n
s
e
.
Cor
r
e
s
pon
di
n
g
Au
t
h
or
:
L
i
l
i
a
na
,
D
e
pa
r
t
m
e
nt
of
Info
rm
a
t
i
c
s
,
P
e
t
ra
Chr
i
s
t
i
a
n
U
ni
ve
rs
i
t
y
,
S
i
w
a
l
a
nke
rt
o
121
-
131
s
t
re
e
t
,
S
u
ra
b
a
ya
60
236,
In
done
s
i
a
.
E
m
a
i
l
:
l
i
l
i
a
n
@
pe
t
ra
.
a
c
.
i
d
1.
I
N
TR
O
D
U
C
TI
O
N
U
nt
i
l
re
c
e
n
t
l
y
,
h
a
nd
ge
s
t
ure
re
c
ogn
i
t
i
on
h
a
s
be
e
n
de
v
e
l
op
e
d
for
va
r
i
ous
p
urpos
e
s
,
s
uc
h
a
s
s
i
gn
l
a
ngu
a
g
e
un
de
rs
t
a
nd
i
ng
[1
-
5]
,
H
u
m
a
n
-
Co
m
put
e
r
Int
e
ra
c
t
i
o
n
[6]
,
v
i
rt
u
a
l
e
nv
i
ron
m
e
nt
i
n
t
e
r
a
c
t
i
on
[4
,
7
-
9]
a
nd
c
ont
rol
l
i
ng
us
i
n
g
rob
ot
[
10
-
12]
.
S
o
m
e
a
p
pl
i
c
a
t
i
ons
us
i
ng
ha
nd
g
e
s
t
ur
e
s
a
s
na
v
i
g
a
t
ors
t
o
w
a
l
k
t
h
rough
a
v
i
rt
u
a
l
e
n
vi
ron
m
e
nt
[7
,
8]
,
v
i
rt
ua
l
k
e
ybo
a
rd
,
c
ont
r
ol
l
e
r
a
p
pl
i
a
nc
e
s
or
de
vi
c
e
i
ns
i
d
e
a
c
e
rt
a
i
n
s
pa
c
e
[10
-
12
],
c
ont
r
ol
l
e
r
robo
t
s
ur
ge
ry
a
nd
us
e
d
i
n
m
e
di
c
a
l
purpos
e
s
a
ppl
i
c
a
t
i
on
s
u
c
h
a
s
phys
i
c
a
l
t
re
a
t
m
e
n
t
[6].
By
t
a
ki
ng
a
dv
a
nt
a
g
e
of
c
urre
nt
vi
r
t
ua
l
re
a
l
i
t
y
(V
R)
t
e
c
hn
ol
og
i
e
s
d
e
v
e
l
op
m
e
nt
,
m
a
ny
a
pp
l
i
c
a
t
i
ons
t
h
a
t
e
nha
nc
e
hum
a
n
l
i
fe
,
i
n
c
l
ud
i
ng
m
e
d
i
c
a
l
pur
pos
e
s
a
p
pl
i
c
a
t
i
o
n
,
ha
v
e
be
e
n
de
ve
l
ope
d
a
s
w
e
l
l
[13
-
15
]
.
U
s
ua
l
l
y
,
a
ft
e
r
i
n
j
ury
a
nd
a
f
t
e
r
s
t
roke
pa
t
i
e
nt
s
n
e
e
d
phys
i
c
a
l
t
re
a
t
m
e
n
t
s
uc
h
a
s
h
a
nd
a
nd
l
e
g
m
o
t
i
on
e
x
e
rc
i
s
e
s
.
O
n
t
h
e
ot
he
r
h
a
nd
,
V
R
t
e
c
hno
l
ogi
e
s
provi
de
a
p
ow
e
rfu
l
hu
m
a
n
-
i
nt
e
rf
a
c
e
i
n
t
e
r
a
c
t
i
on
[1
4
-
15]
a
nd
a
ud
i
ov
i
s
ua
l
fe
e
db
a
c
k
s
i
m
ul
a
t
i
on
[13
,
15
],
a
l
l
ow
c
r
e
a
t
i
n
g
ne
w
e
xe
rc
i
s
e
s
e
a
s
i
l
y
a
nd
s
e
t
t
i
ng
t
he
v
i
rt
u
a
l
e
nv
i
ron
m
e
n
t
fl
e
x
i
bl
y
[13
,
15
-
16
].
S
o
m
e
re
s
e
a
r
c
he
s
pr
ove
d
t
h
a
t
t
he
r
a
p
i
e
s
s
upport
e
d
b
y
V
R
t
e
c
h
nol
o
gi
e
s
c
a
n
i
m
prov
e
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
1693
-
6930
T
E
L
K
O
M
N
IK
A
T
e
l
e
c
om
m
un
Co
m
put
E
l
Con
t
rol
,
V
ol
.
18
,
N
o.
2
,
A
pri
l
2
020:
761
-
7
69
762
m
obi
l
i
t
y
[16
-
19
]
a
nd
V
R
i
nt
e
rfa
c
e
c
a
n
s
i
m
ul
a
t
e
t
h
e
br
a
i
n
b
e
t
t
e
r
[14]
.
E
ve
n
t
houg
h
t
h
e
f
i
rs
t
ge
ne
r
a
t
i
on
of
V
R
s
e
ns
ory
d
e
vi
c
e
s
c
ons
i
d
e
re
d
a
l
a
c
k
of
ha
p
t
i
c
f
e
e
db
a
c
k
[18]
,
n
ow
a
da
ys
,
m
a
ny
c
o
m
pa
ni
e
s
pro
m
i
s
e
fa
s
t
,
a
c
c
ura
t
e
a
nd
pow
e
rful
d
e
vi
c
e
s
[1
9].
M
or
e
ov
e
r,
V
R
d
e
v
i
c
e
s
a
re
c
on
s
i
de
r
e
d
l
ow
-
c
os
t
d
e
vi
c
e
s
[13
,
17]
a
nd
ri
c
h
d
a
t
a
c
ol
l
e
c
t
i
on
r
e
t
ri
e
ve
r
[15
-
16].
In
ph
ys
i
c
a
l
t
he
r
a
py
,
t
h
e
ra
p
i
s
t
s
w
i
l
l
de
s
i
gn
s
e
v
e
ra
l
s
p
e
c
i
fi
c
pa
t
t
e
r
ns
o
f
m
o
t
i
on
s
hou
l
d
be
e
xe
rc
i
s
e
d
by
t
he
p
a
t
i
e
n
t
.
V
R
-
b
a
s
e
d
phys
i
c
a
l
r
e
h
a
bi
l
i
t
a
t
i
on
e
qui
ppe
d
w
i
t
h
m
ot
i
on
s
e
ns
or
(s
)
t
o
s
e
ns
e
ha
nd
o
r
l
e
g
m
o
t
i
on
pe
rfor
m
e
d
by
a
p
a
t
i
e
n
t
.
T
h
e
a
pp
l
i
c
a
t
i
on
ne
e
ds
t
o
fi
nd
ou
t
i
f
t
h
e
m
ot
i
on
i
s
i
n
a
c
c
or
da
n
c
e
w
i
t
h
t
he
d
e
s
i
gn
e
d
pa
t
t
e
rn
of
m
ot
i
on
[14
].
T
he
r
e
s
ul
t
o
f
c
h
e
c
k
i
ng
t
h
e
c
or
re
c
t
ge
s
t
ure
w
i
l
l
b
e
a
r
e
s
pons
e
t
ow
a
rd
t
he
v
i
rt
u
a
l
e
nvi
r
onm
e
nt
[13
,
15]
.
T
he
re
a
r
e
t
w
o
ki
nds
of
m
o
t
i
o
n
s
e
ns
ors
,
w
e
a
r
a
bl
e
s
e
ns
ors
a
nd
c
a
m
e
ra
-
ba
s
e
d
s
e
ns
ors
.
In
c
a
s
e
us
i
ng
a
c
a
m
e
r
a
-
b
a
s
e
d
s
e
ns
or,
fro
m
fr
a
m
e
s
c
a
pt
u
re
d
ove
r
t
i
m
e
,
t
he
d
i
s
pl
a
c
e
m
e
nt
of
hu
m
a
n
j
oi
n
t
s
pos
i
t
i
on
w
i
l
l
be
c
ons
i
d
e
re
d
a
s
h
um
a
n
m
o
t
i
on
[9
-
11,
20
-
24].
M
o
t
i
on
w
i
t
h
c
e
r
t
a
i
n
pa
t
t
e
rns
w
i
l
l
b
e
u
nde
rs
t
ood
a
s
a
g
e
s
t
ur
e
.
G
e
ne
r
a
l
l
y,
h
a
n
d
ge
s
t
ure
s
w
i
l
l
be
c
a
t
e
gori
z
e
d
a
s
h
a
nd
pos
e
,
ha
n
d
s
e
que
n
c
e
of
m
ove
m
e
nt
or
ha
n
d
t
ra
j
e
c
t
ory
,
a
nd
h
a
nd
c
on
t
i
nuous
m
ov
e
m
e
n
t
[2
,
3,
6,
9].
A
h
a
n
d
pos
e
i
s
c
o
ns
i
d
e
ra
b
l
y
s
i
m
pl
e
,
e
a
s
y
t
o
be
c
a
pt
u
re
d
a
nd
r
e
c
og
ni
z
e
d
b
ut
no
t
m
a
ny
pos
e
s
c
a
n
b
e
re
pre
s
e
nt
e
d
us
i
ng
one
ha
n
d
or
dou
bl
e
ha
n
d
w
i
t
hout
a
m
b
i
gui
t
y
[1
,
6
-
8,
10
,
12]
.
T
ra
j
e
c
t
ory
g
e
s
t
u
re
c
ons
i
s
t
s
of
s
e
v
e
ra
l
di
ff
e
r
e
nt
pos
e
s
t
o
r
e
pr
e
s
e
nt
a
w
ho
l
e
g
e
s
t
ur
e
w
hi
l
e
i
n
c
on
t
i
nu
ous
g
e
s
t
ur
e
,
pos
e
s
a
nd
di
s
p
l
a
c
e
m
e
n
t
pos
i
t
i
o
ns
or
j
us
t
on
e
j
o
i
nt
m
ove
m
e
n
t
a
r
e
c
ons
i
de
r
e
d
one
s
i
ngl
e
g
e
s
t
ur
e
.
H
ow
e
ve
r,
t
r
a
j
e
c
t
ory
a
nd
c
on
t
i
n
uous
m
ov
e
m
e
nt
c
ons
i
s
t
of
s
e
v
e
r
a
l
pos
e
s
,
di
r
e
c
t
i
on
a
nd
ori
e
n
t
a
t
i
o
n
c
h
a
ng
i
ng
[3
,
5,
6,
9,
11
,
20
].
S
i
nc
e
no
dur
a
t
i
on
l
i
m
i
t
a
t
i
on
i
n
pe
rfor
m
i
ng
a
g
e
s
t
ur
e
,
i
t
n
e
e
ds
d
ura
t
i
on
norm
a
l
i
z
a
t
i
o
n.
D
yn
a
m
i
c
T
i
m
e
W
ra
p
pi
ng
(D
T
W
)
[
25]
or
de
fi
ne
a
fi
x
d
a
t
a
s
a
m
pl
i
ng
[9]
c
a
n
b
e
us
e
d
a
s
s
ol
u
t
i
o
ns
for
t
he
dur
a
t
i
on
prob
l
e
m
.
V
a
ri
ous
t
e
c
hn
i
qu
e
s
h
a
v
e
b
e
e
n
d
e
ve
l
op
e
d
ba
s
e
d
o
n
w
h
a
t
ki
n
d
of
ge
s
t
ure
t
o
r
e
c
o
gni
z
e
a
nd
w
h
a
t
k
i
nd
of
da
t
a
go
t
fr
om
t
h
e
s
e
ns
or.
Co
l
or
-
b
a
s
e
d
re
c
ogn
i
z
i
ng
ha
nd
pos
e
t
ry
t
o
und
e
rs
t
a
nd
ha
n
d’
s
h
a
pe
,
c
urv
a
t
u
re
be
t
w
e
e
n
f
i
nge
rs
or
h
ow
m
a
ny
fi
ng
e
rs
ope
ne
d
[6
,
10
,
12]
.
Col
or
-
b
a
s
e
d
d
a
t
a
a
l
l
ow
s
a
l
i
t
t
l
e
num
b
e
r
of
h
a
nd
m
ot
i
on
ge
s
t
ure
s
,
s
u
c
h
a
s
s
w
a
p
t
o
l
e
f
t
or
r
i
gh
t
,
pus
h
a
n
d
pu
l
l
ha
n
d
[4
,
7,
8
,
1
0].
S
uc
h
g
e
s
t
ur
e
s
c
a
n
b
e
us
e
d
t
o
na
vi
g
a
t
e
a
v
a
t
a
rs
i
n
a
v
i
rt
u
a
l
e
nvi
ron
m
e
n
t
[4,
7,
8]
or
c
o
nt
r
ol
de
v
i
c
e
s
i
ns
i
d
e
a
roo
m
[10
-
1
2].
H
ow
e
ve
r
,
t
o
re
c
og
ni
z
e
h
a
nd
m
ove
m
e
nt
,
c
o
l
or
-
b
a
s
e
d
d
a
t
a
i
s
not
e
noug
h.
I
t
ne
e
ds
d
e
pt
h
i
nfor
m
a
t
i
o
n
t
o
e
xt
r
a
c
t
a
f
e
a
t
ure
v
e
c
t
or
from
t
he
pa
l
m
a
r
e
a
.
Y
a
ng
us
e
d
H
M
M
[2
6]
,
M
ol
c
ha
n
ov
us
e
d
H
O
G
[11
],
s
o
m
e
ot
h
e
rs
us
e
d
s
pa
t
i
o
-
t
e
m
po
ra
l
fe
a
t
ure
[20
,
23
,
27
-
29
]
a
nd
ot
h
e
rs
us
e
d
m
o
t
i
on
fe
a
t
ur
e
[9
,
24
,
30].
Y
e
t
,
t
o
r
e
c
ogni
z
e
m
or
e
v
a
ri
o
us
ha
nd
m
o
t
i
on
ge
s
t
ur
e
s
,
s
k
e
l
e
t
on
-
ba
s
e
d
da
t
a
i
s
b
e
t
t
e
r
[3,
2
2,
2
4,
28
].
U
s
i
n
g
i
nfor
m
a
t
i
on
o
f
a
l
l
j
o
i
n
t
s
’
pos
i
t
i
on
i
n
a
h
um
a
n’s
ha
nd
,
pa
l
m
di
re
c
t
i
on,
or
i
e
n
t
a
t
i
o
n,
ro
t
a
t
i
on
w
h
i
l
e
m
ov
i
ng
c
a
n
b
e
c
a
l
c
u
l
a
t
e
d.
T
he
m
a
t
t
e
r
i
n
re
c
ogn
i
z
i
ng
ha
n
d
m
ove
m
e
nt
i
s
d
e
t
e
rm
i
n
i
ng
t
he
be
gi
n
-
e
nd
of
a
g
e
s
t
ur
e
a
nd
t
r
a
ns
form
t
h
e
l
e
n
gt
h
-
va
r
i
ous
d
a
t
a
i
nt
o
a
uni
form
fi
x
-
l
e
ng
t
h
ve
c
t
o
r.
D
e
S
m
e
dt
us
e
d
fi
s
h
e
r
ve
c
t
or
t
o
r
e
pr
e
s
e
nt
ve
c
t
ors
b
e
t
w
e
e
n
22
j
o
i
nt
s
i
n
ha
nd
a
n
d
ha
nd
rot
a
t
i
on
[
3],
L
u
us
e
d
p
a
l
m
di
r
e
c
t
i
on
a
nd
fi
nge
r
t
i
p
a
ng
l
e
a
s
f
e
a
t
ure
[5]
,
Y
a
n
g
us
e
d
t
a
ng
e
nt
i
a
l
a
ngu
l
a
r
c
h
a
nge
ove
r
k
e
yfra
m
e
s
a
s
fe
a
t
ur
e
[26]
,
L
i
u
us
e
d
p
a
l
m
’s
di
s
p
l
a
c
e
m
e
n
t
i
nfo
rm
a
t
i
on
ov
e
r
fra
m
e
s
a
s
f
e
a
t
ur
e
[28]
a
nd
ot
h
e
rs
t
ook
a
s
e
r
i
e
s
of
pa
l
m
pos
i
t
i
on
fro
m
s
e
v
e
ra
l
f
ra
m
e
s
a
s
3D
d
a
t
a
c
l
o
ud
[
29
,
30]
.
U
s
i
ng
c
a
m
e
r
a
-
b
a
s
e
d
s
e
ns
ors
,
s
uc
h
a
s
l
e
a
p
m
ot
i
on
a
nd
H
T
C
V
i
ve
,
w
e
f
a
c
e
s
o
m
e
c
ha
l
l
e
nge
s
i
n
c
l
ud
i
ng
va
ri
o
us
t
i
m
e
d
ura
t
i
on
a
nd
v
a
r
i
ous
ori
e
n
t
a
t
i
on
a
nd
d
i
re
c
t
i
on
pe
rfor
m
i
n
g
e
a
c
h
g
e
s
t
ur
e
.
S
om
e
us
e
rs
p
e
rfor
m
a
ge
s
t
ur
e
fa
s
t
e
r
,
t
he
o
t
he
rs
s
l
ow
e
r
.
T
he
s
e
c
on
d
c
ha
l
l
e
ng
e
,
us
e
rs
do
n’t
a
l
w
a
ys
pos
i
t
i
on
t
h
e
i
r
h
a
nd
fa
c
i
ng
t
he
c
a
m
e
ra
.
T
o
o
ve
r
c
om
e
t
he
unun
i
for
m
t
i
m
e
dura
t
i
on
pro
bl
e
m
,
w
e
a
dop
t
e
d
Y
e
a
nd
Ch
e
ng’s
i
de
a
,
s
a
m
pl
i
ng
a
di
s
t
i
n
c
t
n
um
b
e
r
of
po
i
nt
s
fr
om
e
a
c
h
w
hol
e
h
a
nd
m
ove
m
e
n
t
t
ra
c
ki
n
g
[9,
2
5].
F
ro
m
a
l
l
3D
po
i
nt
s
a
re
t
ra
c
ke
d
duri
ng
a
g
e
s
t
u
re
p
e
rfo
rm
a
nc
e
,
w
e
s
a
m
p
l
e
33
poi
n
t
s
un
i
for
m
l
y
.
T
hos
e
33
po
i
nt
s
w
i
l
l
r
e
pre
s
e
nt
our
w
h
ol
e
s
i
ng
l
e
ge
s
t
ur
e
[
9]
.
V
a
r
i
ous
ori
e
nt
a
t
i
on
a
n
d
d
i
re
c
t
i
on
w
i
l
l
be
e
s
t
i
m
a
t
e
d
us
i
ng
c
o
m
pu
t
e
r
gr
a
phi
c
s
a
ppro
a
c
h.
T
o
a
ns
w
e
r
t
h
e
ne
e
d
for
a
c
c
ura
t
e
a
nd
r
e
a
l
-
t
i
m
e
re
s
po
ns
e
phy
s
i
c
a
l
t
r
e
a
t
m
e
nt
a
ppl
i
c
a
t
i
on
,
our
r
e
s
e
a
r
c
h
propos
e
d
a
p
i
p
e
l
i
ne
t
o
s
e
ns
e
a
nd
t
ra
c
k
ha
n
d
ge
s
t
u
re
s
us
i
ng
h
a
nd
m
ov
e
m
e
nt
s
e
ns
or
a
nd
t
o
un
de
rs
t
a
nd
w
h
a
t
ki
nd
of
t
he
pe
rf
orm
e
d
g
e
s
t
ur
e
a
c
c
ura
t
e
l
y
a
n
d
qui
c
k
l
y.
In
ord
e
r
t
o
g
a
i
n
a
robus
t
ha
nd
ge
s
t
ure
c
l
a
s
s
i
f
i
c
a
t
i
on
a
ppl
i
c
a
t
i
on,
w
e
t
ra
ns
form
e
d
e
a
c
h
g
e
s
t
ur
e
i
n
t
o
bi
n
a
ry
i
m
a
ge
s
a
n
d
t
ra
i
n
t
h
e
m
us
i
ng
our
propos
e
d
de
ns
i
t
y
-
ba
s
e
d
CN
N
.
2.
R
ES
EA
R
C
H
M
ET
H
O
D
W
e
prop
os
e
a
pi
p
e
l
i
ne
c
o
nt
a
i
ns
t
w
o
m
a
i
n
ph
a
s
e
s
,
i
m
a
g
e
of
g
e
s
t
ur
e
r
e
gi
s
t
ra
t
i
on
a
nd
g
e
s
t
ur
e
c
l
a
s
s
i
fi
c
a
t
i
on
a
s
s
how
n
i
n
F
i
gur
e
1.
2.
1
.
D
atas
e
t
F
or
our
da
t
a
s
e
t
,
w
e
c
ol
l
e
c
t
e
d
1
4
k
i
nds
of
g
e
s
t
ur
e
s
d
e
s
i
gn
e
d
by
a
phys
i
ot
h
e
r
a
pi
s
t
a
s
s
e
e
n
i
n
F
i
gur
e
2
.
T
he
s
e
ge
s
t
ure
s
a
re
d
e
s
i
gn
e
d
t
o
he
l
p
pa
t
i
e
nt
s
i
m
pr
ovi
ng
t
h
e
i
r
m
ove
m
e
nt
a
bi
l
i
t
y
gra
dua
l
l
y
.
S
t
a
rt
e
d
fro
m
on
e
t
urn
ri
gi
d
m
o
ve
m
e
n
t
,
c
on
t
i
n
ue
t
o
m
or
e
t
h
a
n
one
m
ov
e
m
e
n
t
.
F
o
r
a
dv
a
nc
e
d
t
re
a
t
m
e
n
t
,
pa
t
i
e
nt
s
w
i
l
l
t
r
y
t
o
fol
l
ow
a
s
m
oot
h
m
ove
m
e
nt
,
s
i
m
p
l
e
a
n
d
t
he
n
m
or
e
c
om
pl
e
x.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
K
O
M
N
IK
A
T
e
l
e
c
om
m
un
Co
m
put
E
l
Con
t
rol
A
r
obus
t
m
e
t
ho
d
f
or
V
R
-
bas
e
d
hand
g
e
s
t
ur
e
r
e
c
ogn
i
t
i
on
us
i
n
g
de
ns
i
t
y
-
bas
e
d
C
NN
(
L
i
l
i
ana
)
763
A
l
l
ou
r
g
e
s
t
ur
e
s
c
o
ns
i
s
t
of
on
e
s
i
ngl
e
s
t
r
oke
,
a
c
ont
i
nuous
m
o
ve
m
e
nt
.
E
a
c
h
ge
s
t
u
re
i
s
un
i
qu
e
,
w
i
t
h
no
s
i
m
i
l
a
r
i
t
y
s
ha
p
e
w
i
t
h
90°
l
e
ft
or
ri
g
ht
rot
a
t
i
on.
W
e
us
e
t
h
e
M
N
IS
T
d
a
t
a
s
e
t
s
t
yl
e
,
s
m
a
l
l
i
m
a
g
e
28x
28
p
i
x
e
l
s
,
c
e
n
t
e
r
e
d
,
bl
a
c
k
b
a
c
kgroun
d
w
i
t
h
w
hi
t
e
for
e
gro
und,
pre
s
e
rve
d
t
he
g
e
s
t
ur
e
s
ha
pe
r
a
t
i
o
[31]
.
O
ur
p
i
pe
l
i
n
e
w
i
l
l
ge
ne
r
a
t
e
a
fro
nt
a
l
2D
bi
n
a
ry
i
m
a
g
e
.
It
m
e
a
ns
t
he
s
ha
p
e
w
i
l
l
not
b
e
s
ke
w
e
d
.
P
a
l
m
pos
i
t
i
on
t
ow
a
rd
f
i
ng
e
r’s
t
i
p
pos
i
t
i
on
a
s
o
ri
e
nt
a
t
i
on
a
nd
p
a
l
m
p
os
i
t
i
on
t
ow
a
rd
t
h
e
us
e
r’s
e
ye
a
s
d
i
re
c
t
i
on
.
W
e
us
e
s
m
a
l
l
re
s
o
l
ut
i
on
i
m
a
g
e
s
be
c
a
us
e
our
g
e
s
t
ur
e
s
h
a
p
e
h
a
s
‘one
-
pi
x
e
l
t
h
i
c
kne
s
s
’
a
n
d
s
pa
rs
e
(ha
s
m
a
ny
‘b
l
a
nk
p
i
xe
l
bl
o
c
k’
on
t
he
ba
c
kgro
und
p
a
r
t
).
Ba
s
e
d
o
n
t
he
s
e
c
ondi
t
i
o
ns
,
e
nl
a
rg
i
ng
t
h
e
i
m
a
ge
re
s
o
l
ut
i
on
w
oul
dn
’t
gi
v
e
m
or
e
d
e
t
a
i
l
i
nfor
m
a
t
i
on
.
F
i
gure
1
.
T
w
o
s
t
a
g
e
s
of
g
e
s
t
ur
e
r
e
c
ogni
t
i
o
n
p
i
pe
l
i
n
e
F
i
gure
2
.
D
e
s
i
gn
e
d
ge
s
t
u
re
l
i
s
t
2.
2
.
P
h
as
e
1:
I
m
age
R
e
g
i
s
tr
a
ti
on
of
G
e
s
tu
r
e
S
h
ap
e
G
i
ve
n
33
3D
poi
n
t
s
i
n
t
he
X
Y
Z
c
oord
i
n
a
t
e
,
t
ra
ns
for
m
a
t
i
on
m
a
t
ri
x
U
V
N
s
hou
l
d
be
c
a
l
c
u
l
a
t
e
d
t
o
fi
nd
t
he
f
i
t
t
e
s
t
p
l
a
n
e
t
o
t
hos
e
3D
po
i
nt
s
.
N
a
xi
s
i
s
d
i
re
c
t
i
on
,
V
a
x
i
s
i
s
or
i
e
nt
a
t
i
on.
F
i
rs
t
,
n
orm
a
l
p
l
a
n
e
or
N
c
a
n
be
obt
a
i
ne
d
b
y
a
ppl
yi
ng
L
i
n
e
a
r
L
e
a
s
t
S
q
ua
r
e
a
nd
Cr
a
m
e
r’s
r
ul
e
[32]
.
G
i
v
e
n
pl
a
ne
e
q
ua
t
i
on
a
x+
by
+
c
z
+
d=
0
,
a
s
s
um
i
n
g
t
h
e
z
c
om
p
one
n
t
i
s
a
l
w
a
ys
o
ne
,
t
he
e
qu
a
t
i
on
b
e
c
om
e
s
a
x
+
by
+
d
=
-
z
.
T
he
m
a
t
r
i
x
of
a
l
l
pl
a
ne
e
qua
t
i
o
ns
go
t
f
rom
N
poi
n
t
s
i
s
s
how
n
i
n
(1)
:
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
1693
-
6930
T
E
L
K
O
M
N
IK
A
T
e
l
e
c
om
m
un
Co
m
put
E
l
Con
t
rol
,
V
ol
.
18
,
N
o.
2
,
A
pri
l
2
020:
761
-
7
69
764
[
1
1
1
2
2
1
⋯
1
]
[
]
=
[
−
1
−
2
⋯
−
]
(1)
a
ppl
y
i
ng
l
i
ne
a
r
l
e
a
s
t
s
qua
re
s
on
(
1),
i
t
w
i
l
l
g
e
t
(2)
.
[
∑
∑
∑
∑
∑
∑
∑
∑
]
[
]
=
−
[
∑
∑
∑
]
(2)
Ca
l
c
ul
a
t
i
ng
t
he
c
e
n
t
ro
i
d
a
nd
s
ubt
ra
c
t
i
ng
a
l
l
3D
poi
n
t
s
w
i
t
h
t
he
c
e
nt
roi
d
,
x
,
y
a
nd
z
c
o
ordi
n
a
t
e
i
n
(2)
a
re
de
fi
n
e
d
re
l
a
t
i
ve
t
o
t
he
c
e
n
t
roi
d
.
T
h
e
n
(2)
c
a
n
be
s
i
m
pl
i
fi
e
d
be
c
o
m
e
(3)
.
[
∑
∑
0
∑
∑
0
0
0
]
[
]
=
−
[
∑
∑
0
]
(3)
If
t
h
e
pl
a
ne
i
s
a
rr
a
ng
e
d
t
o
be
t
h
rough
t
h
e
ori
g
i
n
<
0
,
0,
0>
,
t
he
n
on
e
d
i
m
e
ns
i
o
n
i
n
(
3)
c
a
n
b
e
r
e
m
ove
d
,
w
hi
c
h
re
l
a
t
e
s
t
o
d
.
A
pp
l
y
Cra
m
e
r’s
ru
l
e
on
t
h
a
t
r
e
m
ov
e
d
di
m
e
ns
i
on
m
a
t
r
i
x
gi
v
e
s
s
o
m
e
l
i
ne
a
r
(4
-
6)
.
A
s
s
um
i
n
g
a
xi
s
z
i
s
r
e
m
ove
d,
nor
m
a
l
pl
a
ne
(N
a
xi
s
)
w
i
l
l
be
(7)
.
=
Σ
xx
∗
Σ
yy
−
Σ
xy
∗
Σ
xy
(4)
=
(
Σ
yz
∗
Σ
xy
−
Σ
xz
∗
Σ
yy
)
/
d
e
t
(5)
=
(
Σ
xy
∗
Σ
xz
−
Σ
xx
∗
Σ
yz
)
/
d
e
t
(6)
=
[
,
,
1
]
(7)
T
o
pr
e
ve
nt
f
a
i
l
ur
e
i
n
obt
a
i
ni
ng
nor
m
a
l
pl
a
n
e
,
z
c
o
m
pon
e
nt
s
h
oul
d
b
e
a
s
s
um
e
d
t
o
be
a
non
-
z
e
ro
va
l
ue
.
T
he
s
a
m
e
p
roc
e
s
s
i
s
r
e
pe
a
t
e
d
for
t
he
non
-
z
e
ro
x
c
o
m
pon
e
nt
a
nd
non
-
z
e
r
o
y
c
o
m
pon
e
nt
a
l
s
o
.
U
s
i
ng
N
v
e
c
t
or
from
t
he
b
i
gg
e
s
t
d
e
t
v
a
l
u
e
a
s
di
r
e
c
t
i
on
.
O
ri
e
n
t
a
t
i
on
a
xi
s
c
a
n
b
e
c
a
l
c
u
l
a
t
e
d
b
y
pr
e
di
c
t
i
ng
t
h
e
pro
ba
b
l
e
ori
e
n
t
a
t
i
o
n,
up
.
In
c
a
s
e
z
c
om
pone
nt
i
s
t
h
e
non
-
z
e
ro
va
l
ue
,
y
a
xi
s
<
0
,
1,
0
>
w
i
l
l
be
t
he
pr
oba
b
l
e
o
ri
e
nt
a
t
i
on
a
xi
s
.
T
he
n
a
ppl
y
(8
)
t
o
c
a
l
c
u
l
a
t
e
t
h
e
re
a
l
ori
e
n
t
a
t
i
on
a
x
i
s
,
V
.
=
−
(
∙
|
|
)
∗
N
(8)
A
ft
e
r
f
i
nd
i
ng
t
he
U
V
N
c
oord
i
na
t
e
,
a
t
r
a
ns
f
or
m
a
t
i
on
pro
c
e
s
s
c
a
n
be
d
one
us
i
ng
(
9).
u,
v
a
nd
w
a
r
e
c
oe
ff
i
c
i
e
n
t
s
o
f
U
,
V
,
N
a
xi
s
.
x
,
y
,
a
nd
z
a
re
c
oe
ff
i
c
i
e
n
t
s
on
X
,
Y
,
a
n
d
Z
a
xi
s
.
[
1
]
=
[
1
]
[
1
,
1
1
,
2
2
,
1
2
,
2
1
,
3
0
2
,
3
0
3
,
1
3
,
2
4
,
1
4
,
2
3
,
3
0
4
,
3
1
]
(9)
Cra
m
e
r’s
Ru
l
e
i
s
a
d
e
t
e
rm
i
na
nt
-
b
a
s
e
d
pro
c
e
dure
t
ha
t
i
s
us
e
d
t
o
s
o
l
ve
s
ys
t
e
m
s
of
e
qua
t
i
ons
w
i
t
hou
t
s
ol
vi
ng
a
l
l
un
know
n
v
a
ri
a
bl
e
s
.
Cra
m
e
r’s
Ru
l
e
a
l
l
ow
s
u,
v,
w
di
re
c
t
l
y
c
a
l
c
u
l
a
t
e
d
us
i
ng
t
he
s
e
fo
l
l
ow
i
ng
v
e
c
t
or
e
qua
t
i
o
ns
s
how
n
i
n
(14
).
By
s
o
l
vi
ng
u,
v
a
nd
w
va
r
i
a
b
l
e
s
,
a
l
l
e
v
a
l
u
e
s
on
t
h
e
t
ra
ns
for
m
a
t
i
on
m
a
t
ri
x
on
(9
)
c
a
n
be
o
bt
a
i
n
e
d.
=
⋅
(
×
)
(10)
1
=
⃗
⋅
(
×
)
(11)
2
=
⋅
(
⃗
×
)
(12)
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
K
O
M
N
IK
A
T
e
l
e
c
om
m
un
Co
m
put
E
l
Con
t
rol
A
r
obus
t
m
e
t
ho
d
f
or
V
R
-
bas
e
d
hand
g
e
s
t
ur
e
r
e
c
ogn
i
t
i
on
us
i
n
g
de
ns
i
t
y
-
bas
e
d
C
NN
(
L
i
l
i
ana
)
765
1
=
⋅
(
×
⃗
)
(13)
=
1
,
=
2
,
=
3
(14)
L
e
t
⃗
=
<
1
,
0
,
0
>
a
n
d
us
e
Cra
m
e
r
’s
ru
l
e
t
o
c
a
l
c
u
l
a
t
e
1
,
1
,
1
,
2
1
,
3
.
L
e
t
⃗
=
<
0
,
1
,
0
>
t
o
c
a
l
c
u
l
a
t
e
2
,
1
,
2
,
2
2
,
3
.
L
e
t
⃗
=
<
0
,
0
,
1
>
t
o
c
a
l
c
ul
a
t
e
3
,
1
,
3
,
2
3
,
3
.
F
i
n
a
l
l
y
,
l
e
t
⃗
=
<
0
,
0
,
0
>
−
a
nd
c
a
l
c
u
l
a
t
e
4
,
1
,
4
,
2
4
,
3
.
G
e
t
t
i
ng
a
l
l
t
hos
e
e
v
a
l
u
e
s
,
a
t
ra
n
s
f
orm
a
t
i
on
m
a
t
ri
x
i
s
produ
c
e
d
.
T
o
g
e
ne
ra
t
e
a
c
e
n
t
e
r
e
d
28
x
28
b
i
n
a
ry
i
m
a
g
e
,
a
nor
m
a
l
i
z
a
t
i
on
pro
c
e
s
s
i
s
ne
e
de
d.
F
i
rs
t
,
a
dj
us
t
t
he
r
a
t
i
o
of
t
he
a
c
t
ua
l
s
i
z
e
by
d
i
vi
d
e
d
d
e
s
i
r
e
d
i
m
a
g
e
s
i
z
e
,
28
w
i
t
h
m
a
x
i
m
u
m
v
a
l
u
e
be
t
w
e
e
n
d
i
s
t
a
n
c
e
i
n
U
a
x
i
s
a
nd
di
s
t
a
nc
e
i
n
V
a
xi
s
.
(
15)
i
s
us
e
d
t
o
a
d
j
us
t
m
e
n
t
proc
e
s
s
.
=
28
/
ma
x
(
,
)
(15)
M
ul
t
i
pl
y
i
ng
a
l
l
2D
poi
nt
s
w
i
t
h
t
he
r
a
t
i
o,
fi
nd
i
ng
t
he
c
e
nt
e
r
,
s
ubt
r
a
c
t
i
n
g
w
i
t
h
(c
e
nt
e
r
-
<
14
,
14
>
)
,
de
c
i
m
a
l
i
z
i
ng
fl
o
a
t
i
ng
v
a
l
u
e
s
o
f
2D
m
a
ppe
d
p
oi
n
t
s
i
nt
o
i
nt
e
ge
r
p
i
xe
l
s
p
os
i
t
i
on
w
i
l
l
prod
uc
e
di
s
c
ont
i
nuous
l
i
ne
.
Bi
l
i
ne
a
r
i
nt
e
rpo
l
a
t
i
n
g
n
e
e
d
e
d
t
o
s
m
o
ot
h
e
n
t
h
e
m
.
F
i
gur
e
3
v
i
s
ua
l
i
z
e
s
a
l
l
pro
c
e
s
s
e
s
i
n
t
hi
s
pha
s
e
.
F
i
gure
3
.
F
ro
m
l
e
ft
t
o
ri
g
ht
,
3D
po
i
nt
s
c
a
pt
u
re
d
fro
m
a
s
e
ns
o
r,
c
a
l
c
u
l
a
t
i
n
g
t
he
di
re
c
t
i
on
a
nd
ori
e
n
t
a
t
i
on
a
xi
s
of
t
h
e
fi
t
t
e
s
t
p
l
a
ne
,
2D
s
pa
rs
e
bi
n
a
ry
i
m
a
g
e
,
a
ft
e
r
b
i
l
i
ne
a
r
i
n
t
e
rpol
a
t
i
on
2.
3
.
P
h
as
e
2:
H
an
d
g
e
s
tu
r
e
c
l
as
s
i
fi
c
ati
on
In
t
h
i
s
s
t
a
g
e
,
i
ns
p
i
re
d
by
L
e
N
e
t
-
5
t
h
a
t
ha
d
a
l
re
a
dy
pr
ove
d
i
t
s
s
uc
c
e
s
s
on
t
ra
i
ni
n
g
a
l
ow
r
e
s
ol
ut
i
on,
s
m
a
l
l
s
i
z
e
i
m
a
g
e
d
a
t
a
s
e
t
t
h
a
t
c
o
nt
a
i
n
s
s
i
ngl
e
i
nfor
m
a
t
i
on
a
b
out
s
i
m
p
l
e
s
h
a
pe
s
s
uc
h
a
s
M
N
IS
T
a
nd
E
M
N
IS
T
da
t
a
s
e
t
s
a
s
publ
i
s
he
d
o
n
[31]
,
w
e
pr
opos
e
d
our
de
ns
i
t
y
-
ba
s
e
d
CN
N
a
r
c
hi
t
e
c
t
ur
e
.
T
h
i
s
d
e
ns
i
t
y
-
b
a
s
e
d
CN
N
a
rc
h
i
t
e
c
t
ure
h
a
s
c
ons
i
s
t
e
d
3
l
a
y
e
rs
for
fe
a
t
ur
e
e
x
t
ra
c
t
i
on
a
nd
t
w
o
l
a
y
e
rs
f
or
c
l
a
s
s
i
fi
c
a
t
i
on
a
s
s
e
e
n
i
n
F
i
gur
e
4.
F
i
gure
4
.
D
e
ns
i
t
y
-
ba
s
e
d
CN
N
a
rc
h
i
t
e
c
t
ure
for
h
a
nd
ge
s
t
ure
c
l
a
s
s
i
fi
c
a
t
i
on
O
ur
ge
s
t
ure
i
m
a
g
e
c
ha
r
a
c
t
e
r
i
s
t
i
c
s
a
re
h
a
v
i
ng
m
a
n
y
‘bl
a
n
k
bl
oc
k
pi
x
e
l
’
a
s
b
a
c
kgroun
d
a
nd
‘s
i
n
gl
e
-
pi
x
e
l
t
hi
c
kn
e
s
s
’
f
ore
g
round
.
S
o,
our
bi
na
ry
i
m
a
g
e
s
c
ons
i
de
r
e
d
s
pa
rs
e
i
m
a
g
e
s
.
W
e
ne
e
d
t
o
p
re
v
e
nt
t
hos
e
‘b
l
a
n
k
bl
o
c
k
p
i
xe
l
’
c
o
nt
r
i
but
i
ng
t
o
t
h
e
fe
a
t
ure
m
a
ps
a
n
d
bo
a
s
t
t
he
d
e
ns
e
i
m
a
g
e
bl
o
c
k
t
o
c
on
t
ri
but
e
m
o
re
.
P
yra
m
i
d
k
e
rn
e
l
s
i
z
e
a
ppl
i
e
d
to
t
he
fe
a
t
ur
e
e
xt
ra
c
t
i
on
p
a
rt
s
w
i
l
l
s
ol
ve
t
h
i
s
prob
l
e
m
.
B
i
gge
r
k
e
rne
l
s
i
z
e
on
t
he
f
i
rs
t
l
a
ye
r
a
nd
ge
t
t
i
n
g
s
m
a
l
l
e
r
on
t
he
n
e
xt
l
a
ye
r
.
In
t
he
f
i
rs
t
l
a
y
e
r,
b
i
gg
e
r
bl
a
nk
b
l
oc
k
s
on
t
he
b
a
c
kg
round
c
a
n
b
e
e
l
i
m
i
n
a
t
e
d
us
i
ng
a
b
i
g
ke
rne
l
s
i
z
e
for
t
he
c
onv
ol
u
t
i
on
p
roc
e
s
s
.
A
s
t
he
i
m
a
ge
s
i
z
e
g
e
t
t
i
ng
s
m
a
l
l
e
r
,
w
e
a
p
pl
y
a
s
m
a
l
l
e
r
k
e
rn
e
l
s
i
z
e
for
t
h
e
c
on
vol
u
t
i
on
pro
c
e
s
s
.
B
i
g
s
i
z
e
ke
rn
e
l
on
t
he
fi
rs
t
l
a
y
e
r
w
i
l
l
de
t
e
rm
i
ne
w
h
i
c
h
b
l
oc
k
p
i
xe
l
s
ho
ul
d
c
ont
r
i
but
e
m
o
re
.
N
o
t
f
ol
l
ow
i
ng
L
e
N
e
t
-
5
a
r
c
hi
t
e
c
t
ur
e
w
hi
c
h
us
e
d
m
a
x
-
pool
i
ng
l
a
ye
r,
i
ns
t
e
a
d
of
us
i
n
g
m
a
x
po
ol
i
ng,
w
e
us
e
d
l
a
rge
s
t
ri
de
(s
t
ri
de
=
3)
o
n
t
h
e
f
i
rs
t
l
a
y
e
r
c
onvol
ut
i
on
proc
e
s
s
.
Be
c
a
us
e
m
a
x
-
p
ool
i
ng
w
i
l
l
c
a
us
e
bl
a
nk
b
l
oc
k
p
i
x
e
l
s
ne
a
r
t
he
f
ore
g
round
a
re
c
a
l
c
u
l
a
t
e
d
a
s
for
e
grou
nd
i
n
t
h
e
n
e
xt
l
a
ye
r
.
S
i
n
c
e
our
i
nput
i
m
a
g
e
s
i
z
e
i
s
s
m
a
l
l
,
w
e
a
l
s
o
ne
e
d
a
s
m
a
l
l
m
od
e
l
a
s
w
e
l
l
.
T
o
re
m
ove
s
om
e
no
t
s
i
gni
f
i
c
a
nt
n
ode
s
c
o
m
e
f
rom
‘bl
a
nk
b
l
oc
k
pi
xe
l
’,
a
dr
opout
l
a
ye
r
i
s
a
pp
l
i
e
d.
A
ft
e
r
t
h
a
t
,
t
he
out
pu
t
w
i
l
l
be
f
l
a
t
t
e
d
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
1693
-
6930
T
E
L
K
O
M
N
IK
A
T
e
l
e
c
om
m
un
Co
m
put
E
l
Con
t
rol
,
V
ol
.
18
,
N
o.
2
,
A
pri
l
2
020:
761
-
7
69
766
i
nt
o
128
nod
e
s
of
ful
l
y
c
onn
e
c
t
e
d
l
a
y
e
r.
W
e
us
e
c
ros
s
-
e
nt
ro
py
l
os
s
func
t
i
o
n
b
e
c
a
us
e
w
e
ne
e
d
a
prob
a
bi
l
i
s
t
i
c
re
s
ul
t
.
W
e
us
e
14
node
s
for
t
h
e
fi
n
a
l
l
a
ye
r
a
s
t
h
e
nu
m
be
r
of
g
e
s
t
ur
e
s
h
a
pe
c
l
a
s
s
e
s
.
2.
4
.
VR
ap
p
l
i
c
ati
on
s
c
h
e
me
T
o
i
m
p
l
e
m
e
nt
i
n
V
R
-
ba
s
e
d
a
ppl
i
c
a
t
i
on
,
w
e
d
e
ve
l
op
e
d
a
c
l
i
e
n
t
-
s
e
rv
e
r
n
e
t
w
o
rki
ng
s
c
h
e
m
e
a
s
s
how
n
i
n
F
i
gure
5.
T
he
t
r
a
i
n
i
ng
pa
r
t
i
s
i
m
p
l
e
m
e
n
t
e
d
us
i
ng
p
yt
ho
n
w
i
t
h
K
e
ra
s
.
T
he
s
a
m
e
c
a
pt
ur
i
ng
i
m
a
g
e
i
m
pl
e
m
e
nt
a
t
i
on
a
s
i
n
t
he
c
l
i
e
n
t
pa
r
t
i
s
us
e
d
to
c
a
p
t
ure
t
he
d
a
t
a
s
e
t
i
m
a
ge
s
.
A
f
t
e
r
f
i
ni
s
h
i
ng
t
h
e
t
r
a
i
n
i
n
g,
t
he
w
e
i
gh
t
of
t
ha
t
m
ode
l
w
i
l
l
b
e
s
t
or
e
d
on
t
h
e
s
e
rv
e
r
a
nd
c
a
n
be
a
c
c
e
s
s
e
d
by
t
he
V
R
ga
m
e
Cont
e
nt
.
In
t
he
c
l
i
e
n
t
pa
r
t
,
a
s
t
he
a
ppl
i
c
a
t
i
on
do
i
ng
l
oo
p
s
,
t
h
e
ha
nd
c
ont
ro
l
l
e
r
s
e
ns
or
w
i
l
l
c
a
p
t
ur
e
t
he
us
e
r’s
ha
n
d
m
ove
m
e
nt
a
nd
be
s
e
n
t
t
o
t
h
e
s
e
rv
e
r
.
T
h
e
s
e
rve
r
w
i
l
l
g
e
n
e
ra
t
e
a
2D
bi
na
ry
i
m
a
ge
of
t
he
g
e
s
t
ur
e
,
i
nput
i
t
t
o
t
he
de
ns
i
t
y
-
ba
s
e
d
CN
N
a
nd
ge
t
t
h
e
p
re
d
i
c
t
i
on
.
T
h
e
pre
di
c
t
i
o
n
re
s
u
l
t
w
i
l
l
b
e
s
e
nt
t
o
t
he
c
l
i
e
nt
a
nd
s
how
n
i
n
t
h
e
a
ppl
i
c
a
t
i
on
a
s
a
re
s
pons
e
f
or
t
he
us
e
r.
F
i
gure
5
.
C
l
i
e
nt
-
s
e
rv
e
r
n
e
t
w
o
rki
n
g
s
c
he
m
e
3.
R
ES
U
LTS
A
N
D
A
N
A
LY
S
I
S
In
our
e
xp
e
ri
m
e
n
t
s
,
w
e
us
e
7000
ge
s
t
ure
i
m
a
g
e
s
a
s
t
he
da
t
a
s
e
t
,
4
900
fo
r
t
r
a
i
n
i
ng
a
nd
2100
for
t
e
s
t
i
ng
.
T
he
goa
l
of
our
e
xp
e
ri
m
e
nt
i
s
t
o
m
e
a
s
ur
e
h
ow
fa
r
our
pi
p
e
l
i
n
e
s
ui
t
s
t
h
e
pr
obl
e
m
w
e
l
l
.
S
om
e
e
xc
e
l
l
e
n
c
e
s
of
ou
r
m
ode
l
a
re
pyr
a
m
i
d
s
i
z
e
k
e
rn
e
l
a
pp
l
i
e
d
o
n
CN
N
l
a
ye
rs
t
o
a
vo
i
d
bl
a
nk
bl
o
c
k
pi
x
e
l
c
on
t
ri
b
ut
i
ng
i
n
t
he
n
e
xt
l
a
ye
rs
,
re
m
o
ve
m
a
x
-
poo
l
i
n
g
l
a
y
e
rs
a
nd
r
e
p
l
a
c
e
d
t
h
e
m
w
i
t
h
c
onvo
l
ut
i
on
s
t
r
i
de
3,
us
i
n
g
b
i
na
r
y
i
m
a
ge
s
da
t
a
s
e
t
,
no
t
gra
ys
c
a
l
e
i
m
a
g
e
da
t
a
s
e
t
l
i
ke
L
e
N
e
t
5
.
E
v
a
l
u
a
t
e
how
s
ui
t
a
b
l
e
t
he
num
be
r
of
l
a
y
e
rs
a
nd
n
um
b
e
r
of
f
e
a
t
ure
m
a
ps
of
e
a
c
h
l
a
y
e
r
i
n
t
h
e
CN
N
p
a
rt
.
W
e
r
un
our
m
od
e
l
w
i
t
h
600
e
poc
hs
a
nd
1
28
i
m
a
ge
s
p
e
r
b
a
t
c
h
.
Co
m
pa
r
i
ng
o
ur
m
ode
l
w
i
t
h
L
e
N
e
t
-
5
a
s
be
nc
hm
a
rk
m
o
de
l
,
m
e
a
s
uri
ng
w
h
e
t
h
e
r
us
i
ng
o
ur
pyra
m
i
d
s
i
z
e
k
e
rne
l
be
t
t
e
r
t
h
a
n
s
a
m
e
s
i
z
e
k
e
rne
l
for
a
l
l
c
onvo
l
ut
i
on
l
a
ye
rs
,
us
i
n
g
t
hr
e
e
l
a
y
e
rs
i
n
CN
N
p
a
rt
b
e
t
t
e
r
t
h
a
n
a
d
e
e
p
e
r
m
ode
l
,
us
i
ng
t
h
a
t
num
b
e
r
of
f
e
a
t
ur
e
m
a
ps
on
our
m
ode
l
’s
CN
N
i
s
s
ui
t
a
b
l
e
w
i
t
h
o
ur
prob
l
e
m
w
e
l
l
a
nd
w
h
e
t
he
r
t
he
s
a
m
e
m
o
de
l
runn
i
ng
on
gr
a
ys
c
a
l
e
i
m
a
ge
s
w
i
l
l
m
a
ke
a
di
ff
e
re
nc
e
.
T
o
obt
a
i
n
gra
ys
c
a
l
e
i
m
a
g
e
s
,
w
e
m
o
di
f
i
e
d
our
d
a
t
a
s
e
t
by
bl
urr
i
ng
t
h
e
m
us
i
ng
a
g
a
us
s
i
a
n
b
l
ur
.
Com
p
a
ri
s
on
a
c
c
ur
a
c
y
be
t
w
e
e
n
d
e
ns
i
t
y
-
b
a
s
e
d
CN
N
run
o
n
bi
na
ry
i
m
a
ge
s
,
gr
a
ys
c
a
l
e
i
m
a
ge
s
a
nd
us
i
ng
s
a
m
e
s
i
z
e
ke
rn
e
l
s
,
us
i
ng
d
e
e
p
e
r
l
a
y
e
rs
a
nd
a
ppl
yi
ng
a
fe
w
e
r
num
b
e
r
of
fe
a
t
ure
m
a
ps
a
nd
w
i
t
h
our
b
e
nc
h
m
a
rk
m
ode
l
,
L
e
N
e
t
5
i
s
de
s
c
ri
be
d
i
n
T
a
bl
e
1
.
F
i
gur
e
6
s
how
s
d
e
t
a
i
l
i
nf
orm
a
t
i
on
a
b
out
t
he
e
x
a
c
t
va
l
ue
fr
om
e
poc
h
30
un
t
i
l
60
0
w
i
t
h
30
e
p
oc
hs
i
n
c
re
a
s
e
s
.
T
a
b
l
e
1
.
Co
m
p
a
ri
s
on
a
c
c
ura
c
y
be
t
w
e
e
n
s
e
ve
ra
l
m
ode
l
s
w
i
t
h
de
ns
i
t
y
-
ba
s
e
d
m
od
e
l
l
e
N
e
t
5
d
e
e
p
l
a
y
e
r
m
o
d
e
l
s
a
m
e
s
i
z
e
k
e
rn
e
l
fe
w
e
r
fe
a
t
u
re
m
a
p
s
d
e
n
s
i
t
y
-
b
a
s
e
d
CN
N
g
ra
y
s
c
a
l
e
i
m
a
g
e
i
n
p
u
t
E
p
o
c
h
#
o
f
t
h
e
h
i
g
h
e
s
t
a
c
c
u
ra
c
y
240
90
180
600
360
180
H
i
g
h
e
s
t
a
c
c
u
ra
c
y
0
.
9
6
3
3
3
3
0
.
9
6
8
0
0
0
0
.
9
7
1
3
3
3
0
.
9
5
1
0
0
0
0
.
9
7
7
3
3
3
0
.
9
7
6
0
0
0
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
K
O
M
N
IK
A
T
e
l
e
c
om
m
un
Co
m
put
E
l
Con
t
rol
A
r
obus
t
m
e
t
ho
d
f
or
V
R
-
bas
e
d
hand
g
e
s
t
ur
e
r
e
c
ogn
i
t
i
on
us
i
n
g
de
ns
i
t
y
-
bas
e
d
C
NN
(
L
i
l
i
ana
)
767
F
rom
F
i
gure
6
a
n
d
T
a
b
l
e
1
w
e
c
a
n
s
e
e
t
ha
t
us
i
ng
a
de
e
p
e
r
l
a
y
e
r
(7
l
a
y
e
rs
)
m
od
e
l
,
t
he
hi
g
he
s
t
a
c
c
ura
c
y
a
c
h
i
e
v
e
d
i
n
9
0
e
po
c
hs
.
I
t
c
ons
i
de
r
e
d
t
he
f
a
s
t
e
s
t
proc
e
s
s
b
ut
i
t
di
d
not
g
a
i
n
t
he
h
i
gh
e
s
t
a
c
c
u
ra
c
y.
U
s
i
ng
t
h
e
s
a
m
e
s
i
z
e
k
e
rne
l
on
t
h
e
CN
N
p
a
rt
or
gr
a
ys
c
a
l
e
i
m
a
g
e
s
c
a
n
r
e
a
c
h
hi
gh
a
c
c
ura
c
y
i
n
18
0
e
po
c
hs
.
S
l
ow
e
r
pro
c
e
s
s
,
b
e
t
t
e
r
a
c
c
ura
c
y
bu
t
s
t
i
l
l
l
ow
e
r
t
h
a
n
ours
.
Com
pa
r
e
w
i
t
h
t
hos
e
m
od
e
l
s
a
nd
o
ur
de
ns
i
t
y
-
b
a
s
e
d
CN
N
m
od
e
l
,
L
e
N
e
t
5
w
hi
c
h
us
e
d
m
a
x
-
poo
l
i
ng
l
a
y
e
rs
re
a
c
h
t
he
l
ow
e
s
t
a
c
c
u
ra
c
y.
U
s
i
ng
a
f
e
w
e
r
n
um
b
e
r
of
fe
a
t
u
re
m
a
ps
go
t
t
he
l
ow
e
s
t
a
c
c
ura
c
y
a
m
ong
ot
he
rs
.
F
i
gure
6
.
A
c
c
ur
a
c
c
y
c
o
m
p
a
ri
s
on
a
m
on
g
s
e
ve
r
a
l
m
o
de
l
s
4.
C
O
N
C
LU
S
I
O
N
P
yra
m
i
d
k
e
rne
l
s
i
z
e
w
orks
b
e
t
t
e
r
o
n
bi
n
a
ry
i
m
a
ge
s
t
ha
n
on
g
ra
ys
c
a
l
e
i
m
a
g
e
s
e
v
e
n
t
h
ough
ou
r
m
o
de
l
ne
e
d
e
d
m
o
re
e
p
oc
hs
t
o
g
e
t
hi
gh
e
r
a
c
c
u
ra
c
y.
S
i
n
c
e
b
i
na
r
y
i
m
a
g
e
s
h
a
v
e
‘b
l
a
nk
bl
o
c
k
pi
x
e
l
’
a
nd
‘s
i
ngl
e
-
p
i
xe
l
t
hi
c
kne
s
s
’
c
ha
ra
c
t
e
ri
s
t
i
c
s
,
l
a
ye
rs
w
i
t
h
pyr
a
m
i
d
k
e
rn
e
l
s
i
z
e
a
nd
l
a
rg
e
s
t
ri
d
e
c
onvol
ut
i
on
i
n
t
h
e
f
i
rs
t
l
a
y
e
r
a
c
c
om
m
od
a
t
e
d
b
i
n
a
ry
i
m
a
ge
s
be
t
t
e
r
t
ha
n
m
a
x
pool
i
ng
l
a
ye
r
(L
e
N
e
t
5)
b
e
c
a
us
e
t
h
e
y
pre
v
e
nt
t
h
e
‘
bl
a
nk
b
l
o
c
k
pi
xe
l
’
c
on
t
ri
b
ut
e
s
t
o
t
he
fe
a
t
u
re
m
a
ps
.
O
ur
pi
p
e
l
i
ne
i
s
a
b
l
e
t
o
a
c
hi
e
ve
h
i
ghe
r
a
c
c
ura
c
y
w
i
t
h
m
o
re
e
poc
h
t
h
a
n
o
t
he
r
c
om
p
a
r
e
d
m
od
e
l
s
.
E
ve
n
t
hough
ot
h
e
r
m
od
e
l
s
c
a
n
a
c
h
i
e
v
e
t
he
i
r
hi
gh
e
s
t
a
c
c
ur
a
c
y
b
e
fo
re
30
0
e
poc
hs
but
go
t
t
he
a
c
c
ur
a
c
y
d
e
c
r
e
a
s
e
a
ft
e
r
300
e
po
c
hs
.
W
h
i
l
e
our
m
od
e
l
s
t
i
l
l
go
t
pro
m
i
s
i
ng
i
nc
r
e
a
s
e
a
c
c
ur
a
c
y
a
f
t
e
r
300
e
poc
hs
.
B
i
na
r
y
i
m
a
g
e
s
ve
rs
us
gra
ys
c
a
l
e
or
RBG
i
m
a
g
e
s
i
s
not
t
he
on
l
y
re
a
s
on.
O
ur
p
ropos
e
d
m
od
e
l
s
ui
t
a
bl
e
f
or
s
i
m
p
l
e
v
a
ri
ous
i
nfor
m
a
t
i
o
n
(onl
y
t
w
o
v
a
l
ue
s
)
,
l
e
s
s
de
ns
i
t
y
i
m
a
g
e
,
s
pa
rs
e
do
t
s
,
a
n
d
un
a
m
b
i
guous
c
ont
e
nt
i
m
a
ge
da
t
a
s
e
t
s
.
In
t
hi
s
c
a
s
e
,
t
ra
ns
for
m
e
d
‘dr
a
w
i
ng
i
n
t
h
e
a
i
r’
-
l
i
k
e
ge
s
t
u
re
i
nt
o
2D
i
m
a
ge
s
c
ons
i
de
r
e
d
a
s
a
s
u
i
t
a
bl
e
c
hoi
c
e
.
T
h
e
on
l
y
l
i
m
i
t
a
t
i
on
i
n
ou
r
s
ys
t
e
m
i
s
i
t
s
l
a
c
k
of
s
e
q
ue
n
c
e
i
nfor
m
a
t
i
on
of
t
he
g
e
s
t
ur
e
b
e
c
a
us
e
w
e
t
ra
ns
form
e
d
t
h
e
m
i
nt
o
2D
i
m
a
g
e
s
.
F
or
f
urt
h
e
r
phys
i
c
a
l
t
re
a
t
m
e
n
t
a
pp
l
i
c
a
t
i
on
t
h
a
t
ne
e
ds
t
o
t
ra
i
n
g
e
s
t
ur
e
s
b
a
s
e
d
on
t
h
e
i
r
di
ffe
r
e
nt
ord
e
r
of
g
e
s
t
ur
e
but
c
om
e
out
s
i
m
i
l
a
r
2D
i
m
a
ge
s
m
a
pp
i
ng,
i
npu
t
s
e
que
nc
e
g
e
s
t
ur
e
w
i
l
l
s
ol
v
e
t
h
a
t
m
a
t
t
e
r
b
e
t
t
e
r
t
ha
n
t
he
i
np
ut
i
m
a
g
e
.
O
ur
p
ropos
e
d
ne
t
w
orki
ng
s
c
h
e
m
e
w
i
t
h
g
e
s
t
ur
e
c
l
a
s
s
i
f
i
c
a
t
i
on
pi
pe
l
i
n
e
c
a
n
b
e
us
e
d
ge
n
e
ra
l
l
y
a
s
l
o
ng
i
t
re
c
e
i
v
e
s
3D
poi
nt
s
c
l
oud
a
s
i
n
put
.
T
h
e
s
e
3D
poi
n
t
s
gi
v
e
i
nfor
m
a
t
i
on
a
bou
t
t
h
e
bod
y'
s
j
oi
n
t
m
o
ve
m
e
n
t
.
S
e
ve
ra
l
ge
s
t
ur
e
c
on
t
rol
l
e
rs
fo
r
V
R
s
u
c
h
a
s
l
e
a
p
m
ot
i
on
,
ki
n
e
c
t
a
n
d
H
T
C
vi
v
e
s
u
pport
o
ur
s
ys
t
e
m
w
i
t
h
3D
po
i
nt
i
nfor
m
a
t
i
on
.
L
a
t
e
r
,
a
p
pl
y
i
ng
t
he
t
ra
ns
f
e
r
l
e
a
r
ni
n
g
s
c
h
e
m
e
,
pre
s
e
rve
t
he
w
e
i
g
ht
of
t
h
e
CN
N
pa
rt
a
nd
r
e
t
r
a
i
n
o
nl
y
t
he
fu
l
l
y
c
onn
e
c
t
e
d
l
a
y
e
rs
,
our
de
ns
i
t
y
-
ba
s
e
d
CN
N
w
i
t
h
CN
N
l
a
y
e
rs
us
i
ng
py
ra
m
i
d
k
e
rn
e
l
s
i
z
e
w
i
l
l
b
e
c
om
p
a
t
i
bl
e
w
i
t
h
o
t
he
r
s
i
m
i
l
a
r
d
a
t
a
s
e
t
s
.
A
C
K
N
O
WL
ED
G
E
M
EN
TS
T
hi
s
w
ork
w
a
s
s
uppor
t
e
d
i
n
p
a
rt
by
Ins
t
i
t
u
t
e
f
or
Infor
m
a
t
i
on
a
nd
C
om
m
uni
c
a
t
i
ons
T
e
c
hn
ol
og
y
P
rom
ot
i
on
(IIT
P
)
gr
a
nt
fund
e
d
by
t
he
K
or
e
a
gov
e
rn
m
e
n
t
(M
S
IT
)
(N
o
.
201
8
-
0
-
002
45,
D
e
v
e
l
opm
e
nt
of
pre
ve
nt
i
on
t
e
c
hnol
ogy
a
ga
i
ns
t
A
I
dys
fun
c
t
i
on
i
ndu
c
e
d
by
de
c
e
p
t
i
on
a
t
t
a
c
k)
.
A
nd
i
t
w
a
s
a
l
s
o
i
n
p
a
rt
s
up
por
t
e
d
by
t
he
D
ongs
e
o
U
n
i
ve
rs
i
t
y
R
e
s
e
a
r
c
h
Y
e
a
r.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
1693
-
6930
T
E
L
K
O
M
N
IK
A
T
e
l
e
c
om
m
un
Co
m
put
E
l
Con
t
rol
,
V
ol
.
18
,
N
o.
2
,
A
pri
l
2
020:
761
-
7
69
768
R
EF
ER
EN
C
ES
[
1]
L
.
A
b
r
a
h
a
m
,
A
.
U
r
r
u
,
N
.
N
or
m
a
ni
,
M
.
P
.
W
i
l
k
,
M
.
W
a
l
s
h,
B
.
O
’
F
l
ynn
,
“
H
a
nd
T
r
a
c
k
i
ng
a
nd
G
e
s
t
ur
e
R
e
c
ogn
i
t
i
o
n
U
s
i
ng
L
e
ns
l
e
s
s
S
m
a
r
t
S
e
n
s
o
r
s
,
”
i
n
S
e
ns
or
s
,
vo
l
.
1
8
,
no
.
9,
pp
.
28
34,
A
ugu
s
t
20
18
.
[
2]
H
.
C
h
e
ng
,
L
.
Y
a
n
g
a
nd
Z
.
L
i
u,
"
S
u
r
ve
y
o
n
3D
H
a
nd
G
e
s
t
ur
e
R
e
c
o
gni
t
i
on,
"
i
n
I
E
E
E
T
r
a
ns
a
c
t
i
on
s
o
n
C
i
r
c
u
i
t
s
a
nd
Sy
s
t
e
m
s
f
or
V
i
de
o
T
e
c
hn
ol
o
gy
,
vol
.
26
,
no
.
9
,
pp
.
1659
-
16
73
,
S
e
p
t
e
m
be
r
201
6.
[
3]
Q
.
D
e
S
m
e
dt
,
H
.
W
a
nno
us
a
nd
J
.
V
a
nd
e
bo
r
r
e
,
"
S
k
e
l
e
t
on
-
B
a
s
e
d
D
yna
m
i
c
H
a
nd
G
e
s
t
ur
e
R
e
c
ogn
i
t
i
on
,
"
20
16
I
E
E
E
C
o
nf
e
r
e
nc
e
on
C
om
pu
t
e
r
V
i
s
i
on
and
P
at
t
e
r
n
R
e
c
o
gni
t
i
on
W
o
r
k
s
hop
s
(
C
V
P
R
W
)
,
L
a
s
V
e
ga
s
,
N
V
,
pp.
12
06
-
1214
,
201
6.
[
4]
M
.
G
i
l
l
i
e
s
,
“
W
h
a
t
i
s
M
o
ve
m
e
n
t
I
nt
e
r
a
c
t
i
on
i
n
V
i
r
t
ua
l
R
e
a
l
i
t
y
f
o
r
?
,
”
M
O
C
O
'
16
:
P
r
oc
e
e
di
ngs
of
t
h
e
3
rd
I
nt
e
r
na
t
i
ona
l
Sy
m
pos
i
um
on
M
ov
e
m
e
n
t
and
C
om
p
ut
i
n
g
,
pp
.
1
-
4
,
J
u
l
y
201
6.
[
5]
W
.
L
u
,
Z
.
T
o
ng
a
nd
J
.
C
hu
,
"
D
yn
a
m
i
c
H
a
n
d
G
e
s
t
u
r
e
R
e
c
ogn
i
t
i
o
n
W
i
t
h
L
e
a
p
M
o
t
i
on
C
ont
r
o
l
l
e
r
,
"
I
E
E
E
S
i
gn
al
P
r
oc
e
s
s
i
ng
L
e
t
t
e
r
s
,
v
ol
.
23
,
n
o.
9
,
pp
.
1
188
-
11
92,
S
e
pt
e
m
be
r
2016
.
[
6]
J
.
M
.
P
a
l
a
c
i
os
,
C
.
S
a
gue
s
,
E
.
M
on
t
i
j
a
no
,
S
.
L
l
or
e
n
t
e
,
“
H
u
m
a
n
-
C
o
m
put
e
r
I
nt
e
r
a
c
t
i
on
B
a
s
e
d
o
n
H
a
nd
G
e
s
t
u
r
e
U
s
i
ng
R
G
B
-
D
S
e
ns
o
r
s
,
”
S
e
ns
o
r
s
,
vo
l
.
13
,
no
.
9,
pp
.
118
42
-
1186
0,
S
e
pt
e
m
be
r
20
13
[
7]
C
.
K
h
unda
m
,
"
F
i
r
s
t
pe
r
s
on
m
o
ve
m
e
nt
c
o
nt
r
ol
w
i
t
h
pa
l
m
nor
m
a
l
a
n
d
ha
n
d
g
e
s
t
u
r
e
i
n
t
e
r
a
c
t
i
o
n
i
n
v
i
r
t
ua
l
r
e
a
l
i
t
y,
"
201
5
12
th
I
nt
e
r
n
at
i
on
al
J
oi
nt
C
o
nf
e
r
e
nc
e
on
C
om
p
ut
e
r
S
c
i
e
nc
e
an
d
Sof
t
w
a
r
e
E
n
gi
n
e
e
r
i
n
g
(
J
C
SSE
)
,
S
ongk
hl
a
,
pp.
32
5
-
3
30
,
2
015
.
[
8]
F
.
Z
h
a
ng
,
S
.
C
hu,
R
.
P
a
n
,
N
.
J
i
a
nd
L
.
X
i
,
"
D
oub
l
e
ha
nd
-
ge
s
t
u
r
e
i
n
t
e
r
a
c
t
i
on
f
o
r
w
a
l
k
-
t
hr
o
ugh
i
n
V
R
e
nvi
r
on
m
e
nt
,
"
201
7
I
E
E
E
/
A
C
I
S
16
t
h
I
n
t
e
r
na
t
i
ona
l
C
on
f
e
r
e
nc
e
on
C
om
put
e
r
an
d
I
n
f
or
m
at
i
on
S
c
i
e
nc
e
(
I
C
I
S)
,
W
uh
a
n,
pp.
53
9
-
5
44
,
2
017
.
[
9]
G
ua
ngq
i
Y
e
,
J
.
J
.
C
o
r
s
o
a
nd
G
.
D
.
H
a
g
e
r
,
"
G
e
s
t
ur
e
R
e
c
o
gn
i
t
i
on
U
s
i
ng
3D
A
pp
e
a
r
a
n
c
e
a
n
d
M
o
t
i
on
F
e
a
t
u
r
e
s
,
"
20
04
C
onf
e
r
e
n
c
e
on
C
om
p
ut
e
r
V
i
s
i
on
and
P
a
t
t
e
r
n
R
e
c
og
ni
t
i
on
W
o
r
k
s
ho
p
,
W
a
s
h
i
ng
t
on
,
D
C
,
U
S
A
,
pp
.
160
-
1
66
,
2
004
.
[
10]
D.
L
.
D
i
n
h,
J
.
T
.
K
i
m
,
T
.
S
.
K
i
m
,
“
H
a
nd
ge
s
t
ur
e
R
e
c
og
ni
t
i
on
a
nd
I
nt
e
r
f
a
c
e
vi
a
a
D
e
pt
h
I
m
a
g
i
ng
S
e
ns
or
f
o
r
S
m
a
r
t
H
om
e
A
pp
l
i
a
n
c
e
s
,
”
i
n
E
ne
r
gy
P
r
oc
e
d
i
a
,
vo
l
.
62
,
pp
.
576
-
582
,
201
4
.
[
11]
P
.
M
ol
c
ha
n
ov
,
S
.
G
u
pt
a
,
K
.
K
i
m
a
n
d
K
.
P
u
l
l
i
,
"
M
u
l
t
i
-
s
e
n
s
o
r
s
y
s
t
e
m
f
or
dr
i
ve
r
'
s
ha
nd
-
g
e
s
t
ur
e
r
e
c
og
ni
t
i
on,
"
2
015
11t
h
I
E
E
E
I
n
t
e
r
na
t
i
ona
l
C
on
f
e
r
e
n
c
e
and
W
o
r
k
s
hop
s
on
A
ut
om
at
i
c
F
ac
e
and
G
e
s
t
ur
e
R
e
c
ogn
i
t
i
o
n
(
F
G
)
,
L
j
ubl
j
a
na
,
pp.
1
-
8,
2
015
.
[
12]
J.
P
.
S
on
,
A
.
S
ow
m
ya
,
“
S
i
ng
l
e
-
H
a
n
d
e
d
D
r
i
v
i
ng
S
ys
t
e
m
w
i
t
h
K
i
ne
c
t
,
”
I
n
t
e
r
na
t
i
ona
l
C
o
nf
e
r
e
n
c
e
on
H
um
an
-
C
om
pu
t
e
r
I
nt
e
r
a
c
t
i
on
H
C
I
2
013
,
vol
8
005
,
pp.
6
31
-
639
,
J
u
l
y
201
3.
[
13]
C
.
C
a
m
po
r
e
s
i
,
M
.
K
a
l
l
m
a
nn
a
n
d
J
.
J
.
H
a
n
,
"
V
R
s
ol
ut
i
on
s
f
o
r
i
m
p
r
ov
i
ng
ph
ys
i
c
a
l
t
he
r
a
p
y,
"
201
3
I
E
E
E
V
i
r
t
u
al
R
e
al
i
t
y
(
V
R
)
,
L
a
k
e
B
u
e
na
V
i
s
t
a
,
F
L
,
pp
.
77
-
78
,
2
013
.
[
14]
R.
G
.
L
upu
,
D
.
C
.
I
r
i
m
i
a
,
F
.
U
ng
ur
e
a
nu
,
M
.
S
.
P
o
bor
on
i
uc
,
A
.
M
.
,
“
B
C
I
a
n
d
F
E
S
B
a
s
e
d
T
he
r
a
py
f
or
S
t
r
o
ke
R
e
h
a
b
i
l
i
t
a
t
i
on
U
s
i
ng
V
R
F
a
c
i
l
i
t
i
e
s
,
”
in
W
i
r
e
l
e
s
s
C
om
m
un
i
c
at
i
ons
an
d
M
o
bi
l
e
C
om
p
ut
i
ng
,
vo
l
.
20
18,
n
o.
4
,
A
pr
i
l
2018
.
[
15]
D
.
W
h
i
t
e
,
K
.
B
u
r
di
c
k
,
G
.
F
ul
k
,
J
.
S
e
a
r
l
e
m
a
n
a
nd
J
.
C
a
r
r
o
l
l
,
"
A
v
i
r
t
ua
l
r
e
a
l
i
t
y
a
ppl
i
c
a
t
i
on
f
o
r
s
t
r
o
ke
pa
t
i
e
n
t
r
e
ha
bi
l
i
t
a
t
i
on
,
"
i
n
C
y
be
r
P
s
y
c
h
ol
o
gy
&
B
e
hav
i
o
r
,
vo
l
.
9,
no
.
6,
p
p.
1
081
-
1086
,
D
e
c
e
m
be
r
2
006
.
[
16]
J.
E
.
D
e
ut
s
c
h
,
“
V
i
r
t
u
a
l
r
e
a
l
i
t
y
a
n
d
ga
m
i
ng
s
y
s
t
e
m
s
t
o
i
m
pr
o
ve
w
a
l
k
i
ng
a
n
d
m
ob
i
l
i
t
y
f
or
p
e
op
l
e
w
i
t
h
m
us
c
u
l
os
ke
l
e
t
a
l
a
nd
ne
u
r
o
m
u
s
c
u
l
a
r
c
ond
i
t
i
o
ns
,
”
in
S
t
u
di
e
s
i
n
h
e
al
t
h
t
e
c
h
nol
ogy
an
d
i
n
f
o
r
m
at
i
c
s
,
v
ol
.
145
,
pp
.
8
4
-
9
3,
F
e
bua
r
y
200
9.
[
17]
D
.
C
or
be
t
t
a
,
F
.
I
m
e
r
i
,
R
.
G
a
t
t
i
,
“
R
e
h
a
bi
l
i
t
a
t
i
on
t
ha
t
i
nc
or
p
or
a
t
e
s
v
i
r
t
u
a
l
r
e
a
l
i
t
y
i
s
m
or
e
e
f
f
e
c
t
i
v
e
t
ha
n
s
t
a
nd
a
r
d
r
e
ha
bi
l
i
t
a
t
i
on
f
or
i
m
pr
ovi
ng
w
a
l
k
i
ng
s
pe
e
d,
b
a
l
a
nc
e
a
nd
m
o
bi
l
i
t
y
a
f
t
e
r
s
t
r
ok
e
:
a
s
ys
t
e
m
a
t
i
c
r
e
v
i
e
w
,
”
J
o
ur
n
al
o
f
P
hy
s
i
o
t
he
r
apy
,
vo
l
.
6
1,
no
.
3,
p
p.
11
7
-
1
24,
J
ul
y
2
015
.
[
18]
L
.
S
c
hm
i
d
,
A
.
G
l
ä
s
s
e
l
,
C
.
S
c
h
us
t
e
r
-
A
m
f
t
,
“
T
h
e
r
a
pi
s
t
s
’
P
e
r
s
pe
c
t
i
v
e
on
V
i
r
t
ua
l
R
e
a
l
i
t
y
T
r
a
i
n
i
n
g
i
n
P
a
t
i
e
nt
s
a
f
t
e
r
S
t
r
o
ke
:
A
Q
ua
l
i
t
a
t
i
ve
S
t
udy
R
e
po
r
t
i
n
g
F
o
c
u
s
G
r
ou
p
R
e
s
ul
t
s
f
r
o
m
T
hr
e
e
H
o
s
pi
t
a
l
s
,
”
S
t
r
o
k
e
R
e
s
e
ar
c
h
and
T
r
e
a
t
m
e
n
t
vol
.
2016
,
no
.
2,
pp
.
1
-
12
,
D
e
c
e
m
b
e
r
20
16.
[
19]
“
D
oc
u
m
e
nt
a
t
i
on
”
,
R
e
t
r
i
e
v
e
d
f
r
o
m
ht
t
p
s
:
/
/
de
ve
l
o
pe
r
.
l
e
a
pm
ot
i
on
.
c
o
m
/
doc
um
e
n
t
a
t
i
o
n/
,
1
st
F
e
b
2
020
.
[
20]
P
.
W
a
ng
,
W
.
L
i
,
S
.
L
i
u
,
Z
.
G
a
o
,
C
.
T
a
ng
a
nd
P
.
O
gunb
ona
,
"
L
a
r
g
e
-
s
c
a
l
e
I
s
o
l
a
t
e
d
G
e
s
t
u
r
e
R
e
c
ogn
i
t
i
o
n
u
s
i
ng
C
onvo
l
ut
i
o
na
l
N
e
ur
a
l
N
e
t
w
o
r
k
s
,
"
201
6
23r
d
I
n
t
e
r
na
t
i
ona
l
C
on
f
e
r
e
nc
e
on
P
at
t
e
r
n
R
e
c
ogn
i
t
i
o
n
(
I
C
P
R
)
,
C
a
n
c
un
,
pp.
7
-
12
,
2016
.
[
21]
Y
.
Z
hou
,
G
.
J
i
a
n
g,
Y
.
L
i
n
,
“
A
N
o
ve
l
f
i
nge
r
a
nd
h
a
nd
po
s
e
e
s
t
i
m
a
t
i
on
t
e
c
h
ni
q
ue
f
o
r
r
e
a
l
-
t
i
m
e
h
a
nd
ge
s
t
ur
e
r
e
c
o
gn
i
t
i
on
,
”
P
a
t
t
e
r
n
R
e
c
og
ni
t
i
on
,
v
ol
.
49
,
p
p.
10
2
-
11
4
,
A
u
gus
t
20
15
.
[
22]
M
.
F
.
K
a
s
s
i
m
,
M
.
N
.
H
.
M
ohd
,
“
F
oo
d
I
n
t
a
k
e
G
e
s
t
ur
e
M
o
ni
t
o
r
i
n
g
S
y
s
t
e
m
B
a
s
e
d
-
on
D
e
pt
h
S
e
ns
or
”
,
B
u
l
l
e
t
i
n
o
f
E
l
e
c
t
r
i
c
al
E
ng
i
n
e
e
r
i
ng
a
nd
I
n
f
or
m
at
i
c
s
,
vo
l
.
8
,
no
.
2
,
pp
.
470
-
47
6,
J
une
201
9.
[
23]
M
.
G
a
l
i
n
i
u
m
,
J
.
Y
a
p
r
i
,
J
.
P
u
r
na
m
a
,
“
M
a
r
k
e
r
n
e
s
s
m
o
t
i
o
n
C
a
p
t
u
r
e
f
or
3D
H
u
m
a
n
M
ode
l
A
n
i
m
a
t
i
o
n
U
s
i
ng
D
e
p
t
h
C
a
m
e
r
a
”
,
T
E
L
K
O
M
N
I
K
A
T
e
l
e
c
om
m
un
i
c
a
t
i
on
E
l
e
c
t
r
on
i
c
s
C
om
pu
t
i
ng
an
d
C
on
t
r
ol
,
v
ol
.
1
7,
n
o
.
3
,
p
p.
1300
-
13
09
,
J
une
20
19
.
[
24]
L
.
T
i
a
n,
N
.
M
.
T
h
a
l
m
a
nn
,
D
.
T
ha
l
m
a
n
n,
J
.
Z
he
n
g,
“
N
a
t
u
r
e
G
r
a
s
p
i
ng
by
a
C
a
bl
e
-
d
r
i
ve
n
U
nd
e
r
-
a
c
t
u
a
t
e
d
A
nt
h
r
opo
m
o
r
ph
i
c
R
ob
ot
i
c
H
a
n
d,
”
T
E
L
K
O
M
N
I
K
A
T
e
l
e
c
om
m
u
ni
c
a
t
i
on
E
l
e
c
t
r
o
ni
c
s
C
om
pu
t
i
ng
and
C
on
t
r
o
l
,
v
ol
.
17
,
n
o.
1
,
pp.
1
-
7
,
F
e
bu
a
r
y
2
019
.
[
25]
H
.
C
h
e
ng
,
J
.
L
u
o
a
nd
X
.
C
h
e
n,
"
A
w
i
nd
ow
e
d
dyn
a
m
i
c
t
i
m
e
w
a
r
pi
ng
a
ppr
oa
c
h
f
or
3D
c
o
nt
i
nu
ous
ha
nd
ge
s
t
u
r
e
r
e
c
o
gn
i
t
i
on
,
"
2
014
I
E
E
E
I
n
t
e
r
nat
i
o
nal
C
o
nf
e
r
e
n
c
e
on
M
u
l
t
i
m
e
d
i
a
a
nd
E
x
po
(
I
C
M
E
)
,
C
he
ng
du
,
p
p.
1
-
6
,
2014
.
[
26]
C
.
Y
a
ng
,
D
.
K
.
H
a
n,
H
.
K
o
,
“
C
o
nt
i
nuo
us
h
a
nd
ge
s
t
u
r
e
r
e
c
o
gn
i
t
i
on
b
a
s
e
d
on
t
r
a
j
e
c
t
o
r
y
s
ha
pe
i
nf
or
m
a
t
i
on
(
i
n
p
r
e
s
s
)
,
”
i
n
P
at
t
e
r
n
R
e
c
ogn
i
t
i
on
L
e
t
t
e
r
s
,
v
ol
.
99
,
pp.
3
9
-
4
7,
M
a
y
201
7.
[
27]
O
.
K
öpük
l
ü
,
A
.
G
undu
z
,
N
.
K
o
s
e
,
G
.
R
i
go
l
l
,
“
R
e
a
l
-
t
i
m
e
H
a
n
d
G
e
s
t
ur
e
D
e
t
e
c
t
i
on
a
nd
C
l
a
s
s
i
f
i
c
a
t
i
o
n
U
s
i
ng
C
onvo
l
ut
i
o
na
l
N
e
ur
a
l
N
e
t
w
or
ks
,
”
20
19
14
th
I
E
E
E
I
nt
e
r
n
at
i
ona
l
C
onf
e
r
e
nc
e
o
n
A
ut
om
at
i
c
F
a
c
e
&
G
e
s
t
ur
e
R
e
c
og
ni
t
i
on
(
F
G
201
9)
,
Li
l
l
e
,
F
r
a
nc
e
,
p
p.
1
-
8,
2
019
.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
K
O
M
N
IK
A
T
e
l
e
c
om
m
un
Co
m
put
E
l
Con
t
rol
A
r
obus
t
m
e
t
ho
d
f
or
V
R
-
bas
e
d
hand
g
e
s
t
ur
e
r
e
c
ogn
i
t
i
on
us
i
n
g
de
ns
i
t
y
-
bas
e
d
C
NN
(
L
i
l
i
ana
)
769
[
28]
F
.
L
i
u
,
W
.
Z
e
ng
,
C
.
Y
ua
n,
Q
.
W
a
n
g,
Y
.
W
a
n
g,
“
K
i
ne
c
t
-
ba
s
e
d
H
a
nd
G
e
s
t
ur
e
R
e
c
ogn
i
t
i
on
U
s
i
ng
T
r
a
j
e
c
t
o
r
y
I
nf
or
m
a
t
i
on
,
H
a
n
d
M
ot
i
o
n
D
yn
a
m
i
c
s
a
nd
N
e
ur
a
l
N
e
t
w
o
r
k
s
,
”
A
r
t
i
f
i
c
i
a
l
I
nt
e
l
l
i
ge
n
t
R
e
v
i
e
w
,
vo
l
.
52
,
n
o.
1,
pp.
56
3
–
58
3,
A
pr
i
l
2019
.
[
29]
P
.
M
o
l
c
h
a
nov
,
X
.
Y
a
ng
,
S
.
G
up
t
a
,
K
.
K
i
m
,
S
.
T
y
r
e
e
a
nd
J
.
K
a
ut
z
,
"
O
nl
i
ne
D
e
t
e
c
t
i
o
n
a
n
d
C
l
a
s
s
i
f
i
c
a
t
i
on
o
f
D
yna
m
i
c
H
a
nd
G
e
s
t
u
r
e
s
w
i
t
h
R
e
c
ur
r
e
n
t
3D
C
onv
ol
ut
i
on
a
l
N
e
ur
a
l
N
e
t
w
o
r
ks
,
"
20
16
I
E
E
E
C
on
f
e
r
e
nc
e
on
C
om
p
ut
e
r
V
i
s
i
on
an
d
P
at
t
e
r
n
R
e
c
og
ni
t
i
on
(
C
V
P
R
)
,
L
a
s
V
e
g
a
s
,
N
V
,
pp.
4
207
-
421
5,
20
16
.
[
30]
Ho
-
J
o
on
K
i
m
,
J
.
S
.
L
e
e
a
nd
J
.
P
a
r
k
,
"
D
y
na
m
i
c
h
a
nd
ge
s
t
u
r
e
r
e
c
o
gni
t
i
on
u
s
i
ng
a
C
N
N
m
o
de
l
w
i
t
h
3D
r
e
c
e
p
t
i
ve
f
i
e
l
d
s
,
"
200
8
I
n
t
e
r
nat
i
o
nal
C
o
nf
e
r
e
nc
e
on
N
e
ur
a
l
N
e
t
w
or
k
s
and
S
i
g
nal
P
r
o
c
e
s
s
i
ng
,
N
a
nj
i
n
g,
pp
.
14
-
19
,
2008
.
[
31]
Y
.
L
e
c
un
,
L
.
B
ot
t
ou
,
Y
.
B
e
ng
i
o
a
nd
P
.
H
a
f
f
ne
r
,
"
G
r
a
di
e
n
t
-
b
a
s
e
d
l
e
a
r
n
i
ng
a
pp
l
i
e
d
t
o
do
c
u
m
e
n
t
r
e
c
o
gn
i
t
i
on
,
"
i
n
P
r
oc
e
e
d
i
ng
s
o
f
t
he
I
E
E
E
,
vo
l
.
86
,
no
.
11
,
p
p.
22
78
-
2324
,
N
ov
.
1998
.
[
32]
D
.
E
be
r
l
y
,
“
L
e
a
s
t
S
qua
r
e
s
F
i
t
t
i
ng
o
f
D
a
t
a
b
y
L
i
ne
a
r
o
r
Q
ua
d
r
a
t
i
c
S
t
r
uc
t
ur
e
s
”
,
G
e
o
m
e
t
r
i
c
T
ool
s
,
J
ul
y
1
999
.
BI
O
G
R
A
P
H
I
ES
O
F
A
U
T
H
O
R
S
L
i
l
i
an
a
S
he
r
e
c
e
i
v
e
s
h
e
r
B
.
S
f
r
o
m
S
u
r
a
ba
y
a
U
ni
ve
r
s
i
t
y
,
S
u
r
a
b
a
y
a
,
I
ndon
e
s
i
a
i
n
20
02
a
nd
he
r
M
.
E
ng
f
r
o
m
D
ong
s
e
o
U
n
i
ve
r
s
i
t
y,
B
u
s
a
n
,
S
ou
t
h
K
o
r
e
a
i
n
20
09.
N
o
w
s
h
e
i
s
p
ur
s
ui
ng
he
r
D
oc
t
o
r
a
l
de
gr
e
e
i
n
D
on
gs
e
o
U
n
i
v
e
r
i
t
y
,
B
us
a
n,
S
o
ut
h
K
o
r
e
a
.
S
he
ha
s
be
e
n
w
or
k
i
ng
a
s
L
e
c
t
u
r
e
r
a
t
P
e
t
r
a
C
hr
i
s
t
i
a
n
U
ni
v
e
r
s
i
t
y
s
i
n
c
e
2
003
.
H
e
r
i
n
t
e
r
e
s
t
f
i
e
l
d
s
a
r
e
C
o
m
pu
t
e
r
G
r
a
ph
i
c
s
,
C
om
put
e
r
V
i
s
i
on
a
nd
I
nt
e
l
l
i
ge
n
t
S
ys
t
e
m
s
.
Ji
-
H
u
n
C
h
a
e
H
e
r
e
c
e
i
ve
d
h
i
s
B
.
S
,
M
.
S
i
n
C
om
pu
t
e
r
E
n
gi
n
e
e
r
i
ng
f
r
o
m
K
e
i
m
y
ung
U
ni
ve
r
s
i
t
y
,
D
a
e
gu
,
S
ou
t
h
K
or
e
a
,
i
n
2016
a
nd
2018
r
e
s
p
e
c
t
i
ve
l
y.
H
e
i
s
c
u
r
r
e
nt
l
y
a
r
e
s
e
a
r
c
h
e
ng
i
n
e
e
r
a
t
V
i
r
ne
c
t
r
e
s
e
a
r
c
h
I
ns
t
i
t
ut
e
.
H
i
s
r
e
s
e
a
r
c
h
i
n
t
e
r
e
s
t
s
i
nc
l
ud
e
C
o
m
p
ut
e
r
V
i
s
i
on
,
I
m
a
ge
P
r
oc
e
s
s
i
ng
,
C
om
pu
t
e
r
G
r
a
ph
i
c
s
a
nd
G
e
s
t
u
r
e
R
e
c
ogn
i
t
i
o
n.
Joon
-
Jae
L
e
e
H
e
r
e
c
e
i
ve
d
hi
s
B
.
S
.
,
M
.
S
.
,
a
nd
P
h.
D
.
i
n
E
l
e
c
t
r
oni
c
E
n
gi
ne
e
r
i
ng
f
r
o
m
t
he
K
yung
pook
N
a
t
i
on
a
l
U
ni
ve
r
s
i
t
y
,
D
a
e
gu
,
S
o
ut
h
K
or
e
a
,
i
n
19
86
,
1990
,
a
n
d
19
94,
r
e
s
p
e
c
t
i
v
e
l
y.
F
r
o
m
M
a
r
c
h
19
95
t
o
A
ug
us
t
200
7,
h
e
w
a
s
w
i
t
h
t
he
C
o
m
p
ut
e
r
E
ngi
n
e
e
r
i
ng
f
a
c
ul
t
y
a
t
t
he
D
ong
s
e
o
U
ni
ve
r
s
i
t
y,
B
us
a
n,
S
out
h
K
or
e
a
.
H
e
i
s
c
u
r
r
e
n
t
l
y
a
f
ul
l
pr
of
e
s
s
or
o
f
t
h
e
D
e
pa
r
t
m
e
nt
of
G
a
m
e
M
o
bi
l
e
C
ont
e
nt
s
,
K
e
i
m
y
ung
U
n
i
v
e
r
s
i
t
y
.
H
e
w
a
s
a
v
i
s
i
t
i
n
g
s
c
ho
l
a
r
a
t
t
h
e
G
e
or
g
i
a
I
ns
t
i
t
u
t
e
of
T
e
c
hn
ol
ogy
,
A
t
l
a
nt
a
,
f
r
o
m
199
8
t
o
199
9,
f
und
e
d
by
t
h
e
K
o
r
e
a
S
c
i
e
nc
e
a
nd
E
ng
i
ne
e
r
i
n
g
F
ound
a
t
i
on
(
K
O
S
E
F
)
.
H
e
a
l
s
o
w
o
r
ke
d
f
o
r
P
A
R
M
I
C
o
r
po
r
a
t
i
o
n
a
s
a
r
e
s
e
a
r
c
h
a
n
d
de
v
e
l
op
m
e
n
t
m
a
n
a
ge
r
f
or
1
ye
a
r
f
r
o
m
2000
t
o
200
1.
H
i
s
m
a
i
n
r
e
s
e
a
r
c
h
i
n
t
e
r
e
s
t
s
i
nc
l
ud
e
i
m
a
ge
p
r
oc
e
s
s
i
n
g
,
t
h
r
e
e
-
d
i
m
e
n
s
i
ona
l
c
o
m
p
ut
e
r
vi
s
i
on
,
a
nd
f
i
n
ge
r
pr
i
n
t
r
e
c
ogn
i
t
i
o
n.
B
yu
n
g
-
G
o
ok
L
e
e
H
e
r
e
c
e
i
v
e
d
hi
s
B
.
S
.
i
n
M
a
t
he
m
a
t
i
c
s
f
r
o
m
Y
ons
e
i
U
ni
ve
r
s
i
t
y
,
K
or
e
a
,
i
n
19
87,
a
nd
h
i
s
M
.
S
.
a
n
d
P
h
.
D
.
i
n
A
pp
l
i
e
d
M
a
t
he
m
a
t
i
c
s
f
r
o
m
K
o
r
e
a
A
dv
a
nc
e
d
I
n
s
t
i
t
ut
e
o
f
S
c
i
e
n
c
e
a
nd
T
e
c
hn
ol
o
gy
(
K
A
I
S
T
)
i
n
1
989
a
nd
199
3,
r
e
s
pe
c
t
i
v
e
l
y
.
H
e
w
o
r
ke
d
a
t
t
h
e
D
A
C
O
M
C
or
p.
R
&
D
C
e
nt
e
r
a
s
a
s
e
ni
o
r
e
n
gi
n
e
e
r
f
r
o
m
M
a
r
c
h
1993
t
o
F
e
br
u
a
r
y
1
995
.
H
e
ha
s
be
e
n
w
o
r
ki
ng
a
t
D
ongs
e
o
U
ni
ve
r
s
i
t
y,
K
or
e
a
,
s
i
nc
e
19
95
a
nd
i
s
c
ur
r
e
n
t
l
y
a
f
u
l
l
pr
o
f
e
s
s
o
r
w
i
t
h
t
he
D
i
v
i
s
i
on
of
C
o
m
put
e
r
I
nf
or
m
a
t
i
on
E
ng
i
ne
e
r
i
ng
.
H
i
s
r
e
s
e
a
r
c
h
i
nt
e
r
e
s
t
s
i
n
c
l
ude
c
o
m
put
e
r
gr
a
ph
i
c
s
,
c
o
m
p
ut
e
r
-
a
i
de
d
ge
o
m
e
t
r
i
c
de
s
i
gn,
a
nd
i
m
a
ge
pr
oc
e
s
s
i
ng
.
Evaluation Warning : The document was created with Spire.PDF for Python.