I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
pu
t
er
E
ng
ineering
(
I
J
E
CE
)
Vo
l.
1
6
,
No
.
4
,
A
u
g
u
s
t
20
2
6
,
p
p
.
2
0
8
7
~
2
0
9
5
I
SS
N:
2088
-
8
7
0
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijece.
v
1
6
i
4
.
pp
20
8
7
-
2
0
9
5
2087
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
Sema
ntic
-
a
lig
ned
multimo
da
l huma
n
a
ctivity
rec
o
g
ni
tion usin
g
v
isua
l and a
udio
data
Yee
un
P
a
rk
1
,
J
o
o
nh
o
B
y
un
2
,
Siwo
o
B
y
un
3
1
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
mat
i
o
n
,
E
l
e
c
t
r
i
c
a
l
a
n
d
E
l
e
c
t
r
o
n
i
c
En
g
i
n
e
e
r
i
n
g
,
F
a
c
u
l
t
y
o
f
E
n
g
i
n
e
e
r
i
n
g
,
A
n
y
a
n
g
U
n
i
v
e
r
si
t
y
,
A
n
y
a
n
g
,
S
o
u
t
h
K
o
r
e
a
2
R
e
s
e
a
r
c
h
L
a
b
o
r
a
t
o
r
y
o
f
AI
,
F
a
c
u
l
t
y
o
f
En
g
i
n
e
e
r
i
n
g
,
S
e
o
g
a
n
g
U
n
i
v
e
r
si
t
y
,
S
e
o
u
l
,
S
o
u
t
h
K
o
r
e
a
3
D
e
p
a
r
t
me
n
t
o
f
S
o
f
t
w
a
r
e
,
F
a
c
u
l
t
y
o
f
En
g
i
n
e
e
r
i
n
g
,
A
n
y
a
n
g
U
n
i
v
e
r
si
t
y
,
A
n
y
a
n
g
,
S
o
u
t
h
K
o
r
e
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Feb
5
,
2
0
2
6
R
ev
is
ed
Ap
r
1
4
,
2
0
2
6
Acc
ep
ted
Ap
r
2
6
,
2
0
2
6
Hu
m
a
n
a
c
ti
v
it
y
re
c
o
g
n
it
io
n
(H
AR)
re
q
u
ires
ro
b
u
st
p
e
rfo
rm
a
n
c
e
u
n
d
e
r
h
e
tero
g
e
n
e
o
u
s
se
n
si
n
g
c
o
n
d
it
i
o
n
s
fo
r
p
ra
c
ti
c
a
l
d
e
p
lo
y
m
e
n
t.
Ho
we
v
e
r,
sin
g
le
-
m
o
d
a
li
ty
a
p
p
r
o
a
c
h
e
s
a
re
li
m
it
e
d
i
n
c
a
p
tu
r
in
g
th
e
ric
h
c
o
n
tex
t
u
a
l
in
fo
rm
a
ti
o
n
in
h
e
re
n
t
i
n
c
o
m
p
lex
h
u
m
a
n
b
e
h
a
v
io
rs.
T
h
is
p
a
p
e
r
p
re
se
n
ts
a
se
m
a
n
ti
c
-
a
li
g
n
e
d
m
u
lt
imo
d
a
l
H
AR
fra
m
e
wo
rk
th
a
t
in
teg
ra
tes
v
isu
a
l
a
n
d
a
u
d
io
in
f
o
rm
a
ti
o
n
wit
h
o
u
t
a
ss
u
m
in
g
i
n
sta
n
c
e
-
lev
e
l
sy
n
c
h
r
o
n
iz
a
ti
o
n
.
To
a
d
d
re
ss
d
a
tas
e
t
h
e
tero
g
e
n
e
it
y
,
sa
m
p
les
fro
m
th
e
HMDB5
1
v
id
e
o
d
a
tas
e
t
a
n
d
th
e
ES
C
-
5
0
a
u
d
i
o
d
a
tas
e
t
a
re
a
li
g
n
e
d
b
y
m
a
p
p
in
g
fi
n
e
-
g
ra
i
n
e
d
c
las
se
s
in
to
a
sh
a
re
d
h
ig
h
-
lev
e
l
a
c
ti
v
it
y
lab
e
l
s
p
a
c
e
.
F
o
r
e
a
c
h
m
o
d
a
li
ty
,
Re
sN
e
t
-
18
-
b
a
se
d
m
o
d
e
ls
a
re
train
e
d
i
n
d
e
p
e
n
d
e
n
t
ly
u
sin
g
fra
m
e
-
b
a
se
d
v
is
u
a
l
i
n
p
u
ts
a
n
d
6
4
-
b
i
n
M
e
l
-
sp
e
c
tro
g
ra
m
-
b
a
se
d
a
u
d
io
re
p
re
se
n
tatio
n
s.
Du
ri
n
g
in
fe
re
n
c
e
,
th
e
o
u
tp
u
t
lo
g
it
s
o
f
th
e
two
m
o
d
e
ls
a
re
c
o
m
b
in
e
d
t
h
ro
u
g
h
sc
o
re
-
lev
e
l
we
ig
h
ted
li
n
e
a
r
fu
sio
n
.
E
x
p
e
rime
n
tal
re
su
lt
s
sh
o
w
th
a
t
t
h
e
p
ro
p
o
se
d
m
u
lt
imo
d
a
l
a
p
p
r
o
a
c
h
c
o
n
siste
n
tl
y
o
u
t
p
e
rfo
rm
s
u
n
im
o
d
a
l
b
a
se
li
n
e
s
in
term
s
o
f
a
c
c
u
ra
c
y
a
n
d
M
a
c
ro
-
F
1
sc
o
re
,
wi
t
h
p
a
rti
c
u
larl
y
n
o
tab
le
imp
r
o
v
e
m
e
n
ts
in
a
c
ti
v
it
y
g
r
o
u
p
s
wh
e
re
e
n
v
ir
o
n
m
e
n
tal
c
o
n
tex
t
p
lay
s
a
si
g
n
ifi
c
a
n
t
ro
le.
T
h
e
se
fin
d
in
g
s
i
n
d
ica
te
th
a
t
se
m
a
n
ti
c
-
a
li
g
n
e
d
sc
o
re
-
lev
e
l
fu
si
o
n
c
a
n
imp
ro
v
e
re
c
o
g
n
it
io
n
ro
b
u
stn
e
ss
e
v
e
n
u
n
d
e
r
m
ism
a
tch
e
d
d
a
tas
e
t
c
o
n
d
it
io
n
s.
K
ey
w
o
r
d
s
:
Au
d
io
–
v
is
u
al
an
al
y
s
is
Dee
p
lear
n
in
g
Hu
m
an
ac
tiv
ity
r
ec
o
g
n
itio
n
Mu
ltimo
d
al
lear
n
in
g
Sco
r
e
-
lev
el
f
u
s
io
n
Sem
an
tic
alig
n
m
en
t
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
S
iwo
o
B
y
u
n
Dep
ar
tm
en
t o
f
So
f
twar
e,
Facu
lty
o
f
E
n
g
in
ee
r
in
g
,
An
y
a
n
g
U
n
iv
er
s
ity
2
2
,
Sam
d
u
k
-
r
o
,
3
7
th
s
tr
ee
t,
A
n
y
an
g
-
s
h
i,
Ky
o
n
g
k
i,
So
u
th
Ko
r
ea
E
m
ail:
s
wb
y
u
n
@
an
y
an
g
.
ac
.
k
r
1.
I
NT
RO
D
UCT
I
O
N
Hu
m
an
ac
t
iv
ity
r
ec
o
g
n
it
io
n
(
HA
R
)
h
as
b
ec
o
m
e
a
co
r
e
te
c
h
n
o
lo
g
y
in
a
wid
e
r
an
g
e
o
f
a
p
p
li
ca
tio
n
s
,
in
c
lu
d
in
g
s
m
a
r
t
h
e
al
th
c
ar
e,
s
m
ar
t
h
o
m
es
,
an
d
in
t
el
l
ig
en
t
s
u
r
v
ei
ll
an
c
e
s
y
s
t
em
s
[
1
]
,
[
2
]
.
B
y
au
to
m
at
ic
al
ly
r
ec
o
g
n
iz
in
g
h
u
m
an
b
e
h
av
io
r
s
,
H
A
R
en
ab
le
s
co
n
tex
t
-
aw
a
r
e
s
e
r
v
i
ce
s
s
u
ch
a
s
en
v
ir
o
n
m
en
t
al
co
n
tr
o
l
an
d
ab
n
o
r
m
a
l
ac
t
iv
i
ty
d
et
ec
tio
n
.
I
n
p
r
a
ct
ic
al
d
ep
lo
y
m
en
t
,
h
o
w
ev
er
,
H
A
R
s
y
s
te
m
s
m
u
s
t
p
r
o
v
id
e
n
o
t
o
n
ly
h
ig
h
r
ec
o
g
n
it
io
n
a
cc
u
r
ac
y
b
u
t
a
l
s
o
r
o
b
u
s
t
g
en
er
al
iz
at
io
n
an
d
s
t
ab
l
e
in
f
er
en
ce
u
n
d
er
v
ar
y
in
g
r
ea
l
-
w
o
r
ld
co
n
d
i
tio
n
s
.
E
x
is
tin
g
HAR
s
tu
d
ies
h
av
e
p
r
im
ar
ily
f
o
cu
s
ed
o
n
u
n
im
o
d
al
ap
p
r
o
ac
h
es
b
ased
o
n
in
er
tial
m
ea
s
u
r
em
en
t
u
n
its
(
I
MU
s
)
,
v
is
u
al
d
ata,
o
r
a
u
d
io
s
ig
n
als
[
2
]
.
Alth
o
u
g
h
th
ese
a
p
p
r
o
a
ch
es
ca
n
b
e
ef
f
ec
ti
v
e
in
s
p
ec
if
ic
en
v
ir
o
n
m
en
ts
,
th
ey
o
f
ten
f
ail
to
ca
p
tu
r
e
th
e
f
u
l
l
s
em
an
tic
co
n
tex
t
o
f
d
aily
h
u
m
a
n
ac
ti
v
ities
.
I
MU
-
b
ased
m
eth
o
d
s
ar
e
r
elia
b
le
f
o
r
b
asic
p
h
y
s
ical
m
o
v
em
en
ts
in
wea
r
ab
le
s
ettin
g
s
[
3
]
,
b
u
t
ar
e
lim
ited
f
o
r
b
eh
av
io
r
s
in
wh
ich
e
n
v
ir
o
n
m
en
tal
co
n
tex
t
is
im
p
o
r
tan
t.
Vis
u
al
-
b
ased
HAR
ca
n
ef
f
ec
tiv
ely
lear
n
m
o
tio
n
p
atter
n
s
,
b
u
t
v
is
io
n
-
b
ased
m
et
h
o
d
s
m
ay
b
e
v
u
ln
er
a
b
le
to
illu
m
in
atio
n
ch
an
g
es
an
d
o
cc
lu
s
i
o
n
s
[
4
]
,
[
5
]
.
Au
d
io
-
b
ased
HAR
ca
n
p
r
o
v
i
d
e
s
itu
atio
n
al
cu
es
r
elate
d
to
s
u
r
r
o
u
n
d
in
g
s
o
u
n
d
s
,
b
u
t
is
ea
s
ily
af
f
ec
ted
b
y
b
ac
k
g
r
o
u
n
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
1
6
,
No
.
4
,
Au
g
u
s
t
20
2
6
:
2
0
8
7
-
2095
2088
n
o
is
e
an
d
r
ec
o
r
d
in
g
co
n
d
itio
n
s
[
6
]
–
[
8
]
.
T
h
ese
lim
itatio
n
s
m
ak
e
it
d
if
f
icu
lt
f
o
r
u
n
im
o
d
al
m
o
d
els
to
r
o
b
u
s
tly
r
ec
o
g
n
ize
c
o
m
p
lex
b
eh
av
i
o
r
s
in
r
ea
l
-
wo
r
ld
en
v
ir
o
n
m
e
n
ts
.
T
o
o
v
e
r
co
m
e
t
h
ese
lim
itatio
n
s
,
m
u
ltimo
d
al
f
r
am
ewo
r
k
s
h
av
e
b
ee
n
in
c
r
ea
s
in
g
ly
e
x
p
lo
r
e
d
i
n
HAR
b
y
co
m
b
in
in
g
in
f
o
r
m
atio
n
f
r
o
m
m
o
r
e
th
an
o
n
e
s
en
s
in
g
m
o
d
a
lity
[
9
]
–
[
1
1
]
.
Am
o
n
g
th
e
av
a
ilab
le
m
u
ltimo
d
al
in
teg
r
atio
n
s
tr
ateg
ies,
s
ep
ar
ate
m
o
d
ality
-
s
p
ec
if
ic
r
e
p
r
e
s
en
tatio
n
lear
n
in
g
f
o
llo
wed
b
y
o
u
tp
u
t
-
le
v
el
co
m
b
in
atio
n
h
as
b
ee
n
wid
el
y
ad
o
p
ted
in
m
u
ltimo
d
al
lea
r
n
in
g
f
r
am
ewo
r
k
s
[
9
]
,
[
1
0
]
,
[
1
2
]
.
B
ec
au
s
e
th
is
s
tr
ateg
y
k
ee
p
s
th
e
m
o
d
ality
b
r
an
ch
es
in
d
ep
e
n
d
en
t
an
d
p
e
r
f
o
r
m
s
in
teg
r
atio
n
o
n
ly
at
th
e
d
ec
is
io
n
lev
el,
it
r
em
ain
s
s
tr
u
ctu
r
ally
s
im
p
le
an
d
ad
ap
tab
le
f
o
r
p
r
ac
tical
d
ep
lo
y
m
e
n
t
s
ettin
g
s
[
1
0
]
,
[
1
3
]
.
Ho
wev
er
,
m
a
n
y
m
u
ltimo
d
al
HAR
s
tu
d
ies
ar
e
s
till
d
ev
elo
p
ed
u
n
d
er
r
elativ
ely
co
n
tr
o
lled
s
ettin
g
s
a
n
d
o
f
ten
ass
u
m
e
s
y
n
ch
r
o
n
ize
d
o
r
jo
in
tly
co
lle
cted
in
p
u
ts
[
1
4
]
,
[
1
5
]
.
I
n
p
r
a
ctice
,
s
u
ch
p
er
f
ec
t
alig
n
m
e
n
t
is
o
f
ten
d
if
f
ic
u
lt
to
g
u
ar
an
tee
b
ec
au
s
e
o
f
d
if
f
er
e
n
c
es in
s
am
p
lin
g
r
ates,
m
is
s
in
g
d
ata,
o
r
h
eter
o
g
en
e
o
u
s
d
ata
s
o
u
r
ce
s
.
In
th
is
p
ap
er
,
we
p
r
o
p
o
s
e
a
s
em
an
tic
-
alig
n
ed
m
u
ltimo
d
al
HAR
f
r
am
ewo
r
k
th
at
r
elax
es
th
e
ass
u
m
p
tio
n
o
f
in
s
tan
ce
-
lev
el
s
y
n
ch
r
o
n
izatio
n
.
Hete
r
o
g
e
n
eo
u
s
v
id
e
o
an
d
au
d
io
d
atasets
ar
e
alig
n
ed
b
y
m
ap
p
in
g
f
in
e
-
g
r
ain
ed
class
es
in
to
a
s
h
ar
ed
h
ig
h
-
lev
el
ac
tiv
ity
s
em
an
tic
s
p
ac
e,
en
ab
lin
g
s
co
r
e
-
lev
el
f
u
s
io
n
ac
r
o
s
s
m
o
d
alities
.
C
o
n
v
en
tio
n
al
m
u
ltimo
d
al
HAR
ap
p
r
o
a
ch
es
ty
p
ically
r
ely
o
n
s
y
n
c
h
r
o
n
ized
m
u
ltimo
d
al
in
p
u
ts
[
1
1
]
,
[
1
5
]
.
I
n
co
n
tr
ast,
o
u
r
f
r
am
ewo
r
k
ad
d
r
ess
es
a
m
o
r
e
ch
allen
g
in
g
s
ettin
g
in
wh
i
ch
v
id
e
o
an
d
au
d
io
d
ata
ar
e
in
d
ep
e
n
d
en
tly
co
n
s
tr
u
cted
an
d
alig
n
ed
o
n
ly
at
t
h
e
s
em
an
tic
-
g
r
o
u
p
lev
el.
Acc
o
r
d
i
n
g
ly
,
th
e
p
r
o
p
o
s
ed
m
eth
o
d
s
h
o
u
ld
b
e
i
n
ter
p
r
eted
as
a
p
r
ac
tical
late
-
f
u
s
io
n
s
tr
a
teg
y
f
o
r
s
em
an
tically
alig
n
ed
,
n
o
n
-
s
y
n
ch
r
o
n
ize
d
m
u
ltimo
d
al
co
n
d
itio
n
s
r
ath
e
r
t
h
an
as a
d
ir
ec
t r
e
p
lace
m
en
t f
o
r
f
u
lly
s
y
n
c
h
r
o
n
ized
m
u
ltimo
d
al
HAR s
y
s
tem
s
.
T
h
e
m
ain
co
n
t
r
ib
u
tio
n
s
o
f
th
is
wo
r
k
ar
e
s
u
m
m
a
r
ized
as
f
o
llo
ws.
First,
we
d
es
ig
n
a
m
u
ltimo
d
al
HAR
f
r
am
ewo
r
k
u
s
in
g
v
is
u
al
an
d
au
d
io
d
ata
u
s
in
g
HM
DB
5
1
an
d
E
SC
-
5
0
d
atasets
an
d
d
e
m
o
n
s
tr
ate
im
p
r
o
v
ed
p
er
f
o
r
m
an
ce
o
v
er
u
n
im
o
d
al
b
aselin
es
th
r
o
u
g
h
l
ate
f
u
s
i
o
n
.
Seco
n
d
,
we
ex
am
i
n
e
t
h
e
ap
p
lica
b
ilit
y
o
f
m
u
ltimo
d
al
HAR
u
n
d
er
d
ataset
m
is
m
atch
co
n
d
itio
n
s
b
y
em
p
lo
y
in
g
s
em
an
tic
alig
n
m
en
t
a
cr
o
s
s
h
eter
o
g
en
e
o
u
s
d
ata
s
o
u
r
ce
s
.
T
h
ir
d
,
we
ad
o
p
t
a
f
ix
ed
f
u
s
io
n
s
tr
ateg
y
in
wh
ic
h
v
is
u
al
in
f
o
r
m
atio
n
p
lay
s
a
d
o
m
in
an
t
r
o
le
wh
ile
au
d
io
in
f
o
r
m
atio
n
p
r
o
v
id
es c
o
m
p
lem
en
tar
y
c
o
n
tex
t
u
al
cu
es.
2.
M
E
T
H
O
D
2
.
1
.
Da
t
a
s
et
s
I
n
th
is
s
tu
d
y
,
th
e
HM
DB
5
1
d
ataset
was
u
s
ed
f
o
r
v
id
eo
-
b
ased
HAR
,
an
d
th
e
E
SC
-
5
0
d
ataset
was
em
p
lo
y
ed
f
o
r
a
u
d
io
-
b
ased
r
ec
o
g
n
itio
n
.
HM
DB
5
1
is
a
wid
ely
u
s
ed
b
en
c
h
m
ar
k
v
id
eo
d
ataset
co
n
s
is
tin
g
o
f
5
1
ac
tio
n
class
es
[
1
6
]
.
Fro
m
th
i
s
d
ataset,
a
s
u
b
s
et
o
f
class
es
clo
s
ely
r
elate
d
to
d
aily
h
u
m
an
ac
tiv
ities
was
s
elec
ted
an
d
r
eo
r
g
an
ized
in
to
f
iv
e
h
ig
h
-
lev
el
a
ctiv
ity
g
r
o
u
p
s
.
Vid
eo
s
am
p
les
wer
e
p
r
o
c
ess
ed
b
y
s
am
p
lin
g
r
ep
r
esen
tativ
e
f
r
am
es,
wh
ic
h
wer
e
tr
ea
te
d
as
s
tatic
im
ag
e
s
an
d
u
s
ed
as
i
n
p
u
ts
f
o
r
th
e
im
ag
e
-
b
ased
HAR
m
o
d
el.
All f
r
am
es we
r
e
r
esized
to
a
u
n
if
ied
r
eso
lu
tio
n
an
d
n
o
r
m
alize
d
p
r
io
r
to
t
r
ain
in
g
.
E
SC
-
5
0
is
an
en
v
ir
o
n
m
en
tal
s
o
u
n
d
d
ataset
co
m
p
o
s
ed
o
f
5
0
au
d
io
class
es
[
1
7
]
.
I
n
th
is
wo
r
k
,
o
n
ly
s
o
u
n
d
class
es
r
elev
an
t
to
h
u
m
an
ac
tiv
ity
co
n
tex
ts
wer
e
s
elec
ted
an
d
r
eo
r
g
an
ize
d
in
to
th
e
s
am
e
f
iv
e
h
ig
h
-
lev
el
ac
tiv
ity
g
r
o
u
p
s
.
Au
d
io
s
ig
n
als
wer
e
co
n
v
er
ted
in
to
6
4
-
b
i
n
Me
l
-
s
p
ec
tr
o
g
r
a
m
r
ep
r
esen
tatio
n
s
an
d
u
s
ed
a
s
in
p
u
ts
f
o
r
th
e
c
o
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
(
C
NN)
-
b
ased
au
d
io
m
o
d
el.
T
h
e
u
s
e
o
f
M
el
-
s
p
ec
tr
o
g
r
am
s
h
as
b
ee
n
s
h
o
wn
t
o
b
e
ef
f
ec
tiv
e
f
o
r
en
v
ir
o
n
m
en
tal
s
o
u
n
d
an
d
au
d
io
ev
en
t r
ec
o
g
n
itio
n
task
s
[
6
]
,
[
7
]
.
2
.
2
.
Se
m
a
ntic
a
ct
i
v
it
y
g
r
o
up
m
a
pp
ing
HM
DB
5
1
an
d
E
SC
-
5
0
a
r
e
in
d
ep
en
d
e
n
tly
co
n
s
tr
u
cted
d
atasets
with
d
if
f
er
en
t
class
d
ef
in
itio
n
s
an
d
s
em
an
tic
s
tr
u
ctu
r
es
[
1
6
]
,
[
1
7
]
.
T
o
m
itig
ate
th
is
d
ataset
m
is
m
atch
is
s
u
e,
a
s
em
an
tic
alig
n
m
en
t
s
tr
ateg
y
was
ap
p
lied
b
y
g
r
o
u
p
in
g
f
i
n
e
-
g
r
ai
n
ed
class
es b
ased
o
n
ac
tio
n
ch
ar
ac
ter
is
tics
an
d
en
v
ir
o
n
m
en
tal
co
n
tex
t similar
ity
.
Sp
ec
if
ically
,
s
elec
ted
clas
s
es
f
r
o
m
HM
DB
5
1
an
d
E
SC
-
5
0
wer
e
r
eo
r
g
a
n
ized
in
to
f
iv
e
s
h
ar
ed
h
ig
h
-
lev
e
l
ac
tiv
ity
g
r
o
u
p
s
:
Activ
e,
in
ter
ac
tio
n
,
lo
co
m
o
tio
n
,
o
u
td
o
o
r
,
an
d
r
esti
n
g
.
T
h
is
g
r
o
u
p
in
g
was
i
n
ten
d
ed
as
a
co
ar
s
e
s
em
an
tic
alig
n
m
en
t r
ath
e
r
th
a
n
an
ex
ac
t c
lass
-
lev
el
co
r
r
esp
o
n
d
en
ce
b
etwe
en
HM
DB
5
1
an
d
E
SC
-
50.
T
h
is
s
em
an
tic
g
r
o
u
p
in
g
e
n
ab
les
th
e
p
r
ed
ictio
n
r
esu
lts
f
r
o
m
th
e
v
id
eo
an
d
a
u
d
io
m
o
d
a
liti
es
to
b
e
co
m
b
in
ed
with
in
a
s
h
a
r
ed
ac
tiv
ity
lab
el
s
p
ac
e
with
o
u
t
ass
u
m
in
g
in
s
tan
ce
-
lev
el
s
y
n
ch
r
o
n
izatio
n
.
T
a
b
le
1
p
r
esen
ts
th
e
co
m
p
lete
class
-
lev
el
s
em
an
tic
m
ap
p
in
g
u
s
ed
in
th
is
s
tu
d
y
,
in
clu
d
in
g
t
h
e
r
etai
n
ed
HM
DB
5
1
an
d
E
SC
-
5
0
class
es a
s
s
ig
n
ed
to
ea
ch
ac
tiv
ity
g
r
o
u
p
.
Fo
r
th
e
im
ag
e
m
o
d
ality
,
1
0
s
e
lecte
d
HM
DB
5
1
clas
s
es
wer
e
g
r
o
u
p
e
d
in
to
f
iv
e
ac
tiv
ity
g
r
o
u
p
s
,
with
2
,
2
,
3
,
1
,
an
d
2
class
es
ass
ig
n
ed
to
ac
tiv
e,
in
ter
ac
tio
n
,
lo
c
o
m
o
tio
n
,
o
u
td
o
o
r
,
a
n
d
r
esti
n
g
,
r
esp
ec
tiv
ely
.
Fo
r
th
e
au
d
io
m
o
d
ality
,
1
5
s
elec
ted
E
SC
-
5
0
class
es
wer
e
m
ap
p
ed
to
th
e
s
am
e
g
r
o
u
p
s
,
with
4
,
3
,
2
,
4
,
a
n
d
2
class
es
ass
ig
n
ed
,
r
esp
ec
tiv
ely
.
O
v
er
al
l,
1
0
o
f
5
1
HM
DB
5
1
class
es
an
d
1
5
o
f
5
0
E
SC
-
5
0
class
es
wer
e
r
etain
ed
.
T
h
e
g
r
o
u
p
in
g
was
d
eter
m
in
ed
th
r
o
u
g
h
m
a
n
u
al
s
em
an
tic
in
s
p
ec
ti
o
n
to
e
n
s
u
r
e
cr
o
s
s
-
m
o
d
al
in
ter
p
r
etab
ilit
y
,
an
d
th
e
s
h
ar
ed
f
iv
e
-
g
r
o
u
p
lab
el
s
p
ac
e
was u
s
ed
f
o
r
m
u
ltimo
d
al
p
r
e
d
i
ctio
n
an
d
late
f
u
s
io
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
S
ema
n
tic
-
a
lig
n
ed
mu
ltimo
d
a
l
h
u
ma
n
a
ctivity
r
ec
o
g
n
itio
n
u
s
in
g
visu
a
l
…
(
Yeeu
n
P
a
r
k
)
2089
T
ab
le
1
.
Sem
an
tic
m
a
p
p
in
g
o
f
r
etain
ed
HM
DB
5
1
an
d
E
SC
-
5
0
class
es in
to
f
iv
e
ac
tiv
ity
g
r
o
u
p
s
G
r
o
u
p
R
e
t
a
i
n
e
d
H
M
D
B
5
1
c
l
a
ss
e
s
N
o
.
R
e
t
a
i
n
e
d
ESC
-
5
0
c
l
a
sses
N
o
.
D
e
scri
p
t
i
o
n
A
c
t
i
v
e
j
u
m
p
,
si
t
up
2
c
o
u
g
h
i
n
g
,
sn
e
e
z
i
n
g
,
v
a
c
u
u
m
c
l
e
a
n
e
r
,
w
a
sh
i
n
g
m
a
c
h
i
n
e
4
P
h
y
s
i
c
a
l
e
x
e
r
c
i
se
o
r
i
n
t
e
n
si
v
e
b
o
d
y
mo
v
e
me
n
t
I
n
t
e
r
a
c
t
i
o
n
t
a
l
k
,
e
a
t
2
c
l
a
p
p
i
n
g
,
l
a
u
g
h
i
n
g
,
s
p
e
e
c
h
3
H
u
ma
n
i
n
t
e
r
a
c
t
i
o
n
o
r
o
b
j
e
c
t
-
r
e
l
a
t
e
d
a
c
t
i
v
i
t
y
Lo
c
o
m
o
t
i
o
n
w
a
l
k
,
r
u
n
,
c
l
i
m
b
s
t
a
i
r
s
3
f
o
o
t
st
e
p
s
,
k
e
y
b
o
a
r
d
t
y
p
i
n
g
2
M
o
v
e
me
n
t
a
n
d
d
i
sp
l
a
c
e
m
e
n
t
a
c
t
i
v
i
t
i
e
s
O
u
t
d
o
o
r
c
l
i
mb
1
d
o
o
r
w
o
o
d
k
n
o
c
k
,
d
o
o
r
w
o
o
d
c
r
e
a
k
s,
h
a
mm
e
r
,
c
h
a
i
n
s
a
w
4
A
c
t
i
v
i
t
i
e
s
a
sso
c
i
a
t
e
d
w
i
t
h
o
u
t
d
o
o
r
o
r
e
n
v
i
r
o
n
m
e
n
t
a
l
c
o
n
t
e
x
t
R
e
st
i
n
g
si
t
,
st
a
n
d
2
b
r
e
a
t
h
i
n
g
,
s
n
o
r
i
n
g
2
S
t
a
t
i
c
p
o
st
u
r
e
o
r
l
o
w
-
m
o
v
e
m
e
n
t
s
t
a
t
e
2
.
3
.
Da
t
a
prepro
ce
s
s
ing
Fo
r
th
e
v
is
u
al
m
o
d
ality
,
s
am
p
led
f
r
am
e
im
ag
es
wer
e
r
esi
ze
d
to
1
2
8
×1
2
8
an
d
n
o
r
m
ali
ze
d
b
ef
o
r
e
b
ein
g
u
s
ed
as
m
o
d
el
in
p
u
ts
.
T
o
im
p
r
o
v
e
g
en
er
aliza
tio
n
p
e
r
f
o
r
m
a
n
ce
,
b
asic
im
a
g
e
au
g
m
en
tatio
n
tech
n
i
q
u
es,
in
clu
d
in
g
r
o
tatio
n
a
n
d
h
o
r
izo
n
tal
f
lip
p
in
g
,
wer
e
ap
p
lied
.
Fo
r
th
e
au
d
io
m
o
d
ality
,
6
4
-
b
i
n
Me
l
-
s
p
ec
tr
o
g
r
a
m
s
wer
e
n
o
r
m
alize
d
af
ter
tr
an
s
f
o
r
m
atio
n
.
T
o
en
h
a
n
ce
r
o
b
u
s
tn
ess
ag
ain
s
t
v
ar
iatio
n
s
in
r
ec
o
r
d
in
g
en
v
ir
o
n
m
en
ts
,
ti
m
e
-
an
d
f
r
eq
u
en
cy
-
m
ask
in
g
b
ased
au
g
m
en
tatio
n
was
ap
p
lied
[
1
8
]
.
All
p
r
ep
r
o
ce
s
s
in
g
an
d
au
g
m
en
tatio
n
p
r
o
ce
d
u
r
es
wer
e
co
n
d
u
cted
i
n
d
ep
en
d
en
tly
f
o
r
ea
ch
m
o
d
ality
,
tak
in
g
in
to
ac
c
o
u
n
t t
h
eir
r
esp
ec
tiv
e
ch
a
r
ac
ter
is
tics
.
2
.
4
.
I
ma
g
e
-
ba
s
ed
H
AR
mo
del
T
h
e
im
ag
e
-
b
ased
HAR
m
o
d
el
was
d
esig
n
ed
u
s
in
g
a
2
D
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
with
R
esNet
-
1
8
as
th
e
b
ac
k
b
o
n
e
a
r
ch
itectu
r
e.
R
esNet
-
1
8
p
r
o
v
id
es
ef
f
i
cien
t
p
er
f
o
r
m
an
ce
with
a
r
el
ativ
ely
lig
h
tweig
h
t
s
tr
u
ctu
r
e
wh
ile
m
ain
tain
in
g
tr
ain
in
g
s
tab
ilit
y
th
r
o
u
g
h
r
esid
u
al
co
n
n
ec
tio
n
s
[
1
9
]
.
I
n
p
u
t
im
a
g
es
s
am
p
led
f
r
o
m
v
id
eo
f
r
a
m
es
wer
e
p
r
o
ce
s
s
ed
b
y
a
p
r
e
-
tr
ain
ed
R
esNet
-
1
8
n
etwo
r
k
to
ex
tr
ac
t
s
p
atial
f
ea
tu
r
es.
T
h
e
e
x
tr
ac
ted
f
ea
tu
r
es
wer
e
p
ass
ed
th
r
o
u
g
h
f
u
lly
co
n
n
ec
ted
la
y
er
s
to
p
r
o
d
u
ce
class
-
wis
e
lo
g
its
co
r
r
esp
o
n
d
in
g
t
o
th
e
p
r
ed
ef
in
e
d
ac
tiv
ity
g
r
o
u
p
s
.
R
ath
er
th
a
n
ex
p
licitly
m
o
d
elin
g
tem
p
o
r
al
r
elatio
n
s
h
ip
s
,
th
is
s
tu
d
y
ad
o
p
ted
a
f
r
am
e
-
b
ased
s
am
p
lin
g
ap
p
r
o
ac
h
to
r
e
d
u
ce
co
m
p
u
tatio
n
al
co
m
p
lex
ity
wh
ile
p
r
eser
v
in
g
r
e
p
r
esen
tativ
e
v
i
s
u
al
p
atter
n
s
[
2
0
]
.
T
h
is
d
esig
n
is
also
co
n
s
is
ten
t
w
i
th
ef
f
icien
t
v
id
eo
u
n
d
er
s
tan
d
i
n
g
f
r
am
ewo
r
k
s
in
t
en
d
ed
f
o
r
p
r
ac
tical
d
ep
lo
y
m
en
t
[
2
1
]
.
2
.
5
.
Audi
o
-
ba
s
ed
H
AR
m
o
d
el
T
h
e
au
d
io
-
b
ased
HAR
m
o
d
e
l
was
also
co
n
s
tr
u
cted
u
s
in
g
R
esNet
-
1
8
ar
ch
itectu
r
e.
Au
d
io
s
ig
n
als
wer
e
co
n
v
e
r
ted
i
n
to
6
4
-
b
in
Me
l
-
s
p
ec
tr
o
g
r
am
r
ep
r
esen
tati
o
n
s
an
d
tr
ea
ted
as
2
D
in
p
u
ts
to
th
e
C
NN.
Me
l
-
s
p
ec
tr
o
g
r
am
s
ef
f
ec
tiv
ely
ca
p
tu
r
e
tim
e
-
f
r
eq
u
e
n
cy
p
atter
n
s
a
n
d
h
a
v
e
b
ee
n
wid
ely
u
s
ed
i
n
e
n
v
ir
o
n
m
en
tal
s
o
u
n
d
an
d
au
d
io
e
v
en
t
r
ec
o
g
n
itio
n
task
s
[
6
]
,
[
7
]
.
T
h
e
s
p
ec
tr
o
g
r
am
in
p
u
ts
wer
e
p
r
o
ce
s
s
ed
th
r
o
u
g
h
th
e
C
NN
to
ex
tr
ac
t
d
is
cr
im
in
ativ
e
f
ea
tu
r
e
s
,
f
o
llo
wed
b
y
f
u
lly
co
n
n
ec
t
ed
lay
er
s
to
o
u
tp
u
t
class
-
wis
e
lo
g
its
.
Usi
n
g
an
id
en
tical
n
etwo
r
k
a
r
ch
itectu
r
e
f
o
r
b
o
th
m
o
d
alities
h
elp
ed
m
ai
n
tain
co
n
s
is
ten
cy
b
etwe
en
t
h
e
two
b
r
an
ch
es
.
2
.
6
.
Sco
re
-
lev
el
la
t
e
f
us
io
n s
t
ra
t
eg
y
T
h
e
p
r
e
d
ictio
n
r
esu
lts
f
r
o
m
t
h
e
im
ag
e
-
b
ased
an
d
au
d
io
-
b
a
s
ed
m
o
d
els
wer
e
co
m
b
in
ed
u
s
in
g
a
late
f
u
s
io
n
s
tr
ateg
y
.
I
n
th
is
ap
p
r
o
ac
h
,
ea
ch
m
o
d
ality
was
tr
ain
e
d
in
d
e
p
en
d
e
n
tly
,
a
n
d
th
ei
r
o
u
tp
u
ts
wer
e
f
u
s
ed
at
th
e
in
f
er
en
ce
s
tag
e
[
1
2
]
.
L
at
e
f
u
s
io
n
o
f
f
er
s
a
s
im
p
le
an
d
f
lex
ib
le
s
tr
u
ctu
r
e
an
d
d
o
e
s
n
o
t
r
eq
u
ir
e
s
tr
ict
alig
n
m
en
t b
etwe
en
m
o
d
alities
[
1
0
]
,
[
1
3
]
.
L
et
an
d
d
en
o
te
th
e
lo
g
its
g
en
er
ated
b
y
t
h
e
im
ag
e
-
b
a
s
ed
an
d
au
d
io
-
b
ased
m
o
d
els
,
r
esp
ec
tiv
ely
.
T
h
e
f
u
s
ed
lo
g
its
wer
e
co
m
p
u
te
d
u
s
in
g
a
weig
h
t
ed
lin
ea
r
s
u
m
,
as d
e
f
in
ed
i
n
(
1
)
:
=
0
.
6
+
0
.
4
(
1
)
T
h
is
f
u
s
io
n
s
tr
ateg
y
is
d
ef
i
n
ed
as
s
co
r
e
-
lev
el
f
u
s
io
n
,
as
it
co
m
b
in
es
p
r
e
d
ictio
n
s
co
r
es
with
i
n
a
s
h
ar
e
d
ac
tiv
ity
lab
el
s
p
ac
e
with
o
u
t
as
s
u
m
in
g
in
s
tan
ce
-
lev
el
s
y
n
ch
r
o
n
izatio
n
.
A
So
f
tMa
x
f
u
n
ctio
n
was
ap
p
lied
to
th
e
f
u
s
ed
lo
g
its
to
o
b
tain
f
in
al
cl
ass
p
r
o
b
ab
ilit
ies,
an
d
th
e
clas
s
with
th
e
h
ig
h
est
p
r
o
b
ab
ilit
y
was
s
elec
ted
as
th
e
f
in
al
p
r
ed
ictio
n
.
2
.
7
.
E
x
perim
ent
a
l set
up
a
nd
ev
a
lua
t
io
n m
e
t
rics
Fig
u
r
e
1
illu
s
tr
ates
th
e
e
x
p
e
r
im
en
tal
s
etu
p
o
f
th
e
p
r
o
p
o
s
ed
m
u
ltimo
d
al
HAR
f
r
am
e
wo
r
k
.
All
ex
p
er
im
en
ts
wer
e
co
n
d
u
cted
u
s
in
g
th
e
Py
T
o
r
c
h
d
ee
p
le
ar
n
in
g
f
r
am
ewo
r
k
u
n
d
e
r
a
u
n
if
ied
ex
p
er
im
en
tal
en
v
ir
o
n
m
en
t.
B
o
th
im
a
g
e
-
b
a
s
ed
an
d
au
d
io
-
b
ased
m
o
d
els
wer
e
im
p
lem
en
ted
u
s
in
g
th
e
s
am
e
R
esNet
-
1
8
ar
ch
itectu
r
e,
with
m
o
d
ality
-
s
p
ec
if
ic
p
r
e
p
r
o
ce
s
s
in
g
a
p
p
lied
to
th
e
in
p
u
t
d
ata.
I
m
a
g
eNe
t
p
r
e
-
tr
ain
ed
weig
h
ts
wer
e
u
s
ed
f
o
r
m
o
d
el
i
n
itializatio
n
,
f
o
llo
we
d
b
y
f
in
e
-
tu
n
in
g
[
1
9
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
1
6
,
No
.
4
,
Au
g
u
s
t
20
2
6
:
2
0
8
7
-
2095
2090
Fig
u
r
e
1
.
E
x
p
er
im
e
n
tal
s
etu
p
o
f
th
e
p
r
o
p
o
s
ed
m
u
ltimo
d
al
HAR f
r
am
ewo
r
k
T
h
e
im
ag
e
-
b
ased
m
o
d
el
u
s
ed
n
o
r
m
alize
d
in
p
u
ts
with
a
r
es
o
lu
tio
n
o
f
1
2
8
×
1
2
8
p
ix
els
an
d
s
tan
d
ar
d
au
g
m
en
tatio
n
tech
n
i
q
u
es.
T
h
e
au
d
io
-
b
ased
m
o
d
el
em
p
lo
y
ed
6
4
-
b
in
Me
l
-
s
p
ec
tr
o
g
r
am
i
n
p
u
ts
d
e
r
iv
ed
f
r
o
m
E
SC
-
5
0
d
ata,
with
tim
e
-
an
d
f
r
eq
u
en
c
y
-
m
ask
in
g
b
ased
au
g
m
en
tatio
n
ap
p
lied
to
im
p
r
o
v
e
r
o
b
u
s
tn
ess
[
1
8
]
.
All
m
o
d
els we
r
e
d
esig
n
ed
to
s
h
ar
e
an
id
en
tical
o
u
tp
u
t sp
ac
e,
e
n
a
b
lin
g
d
ir
ec
t
f
u
s
io
n
d
u
r
in
g
in
f
e
r
en
ce
.
Fo
r
ea
ch
m
o
d
ality
,
s
am
p
les
b
elo
n
g
in
g
to
th
e
r
etain
e
d
class
es
wer
e
r
an
d
o
m
l
y
d
iv
id
e
d
in
t
o
tr
ain
in
g
an
d
v
alid
atio
n
s
u
b
s
ets
with
a
r
atio
o
f
8
:2
.
T
h
e
m
o
d
els
wer
e
t
r
ain
ed
f
o
r
8
ep
o
c
h
s
with
a
b
at
ch
s
ize
o
f
1
6
a
n
d
a
lear
n
in
g
r
ate
o
f
8
×1
0
-
5
.
All
r
ep
o
r
ted
r
esu
lts
wer
e
o
b
tain
ed
u
n
d
er
a
s
in
g
le
f
ix
ed
ex
p
er
i
m
en
tal
s
ettin
g
an
d
th
er
ef
o
r
e
r
ep
r
esen
t
o
n
e
f
in
al
r
u
n
r
ath
er
t
h
an
av
e
r
ag
es
o
v
er
r
ep
ea
te
d
tr
ials
.
Acc
o
r
d
i
n
g
ly
,
th
e
r
ep
o
r
ted
p
er
f
o
r
m
an
ce
v
alu
es
s
h
o
u
ld
b
e
in
ter
p
r
eted
as
ex
p
lo
r
ato
r
y
ev
id
en
ce
o
f
th
e
f
ea
s
ib
ilit
y
o
f
s
em
an
tics
-
alig
n
ed
m
u
ltimo
d
al
f
u
s
io
n
u
n
d
er
n
o
n
-
in
s
tan
ce
-
s
y
n
ch
r
o
n
ized
d
ata
c
o
n
d
itio
n
s
,
r
ath
er
th
a
n
as
s
tati
s
tically
ex
h
au
s
tiv
e
r
esu
lts
ac
r
o
s
s
m
u
ltip
le
r
ep
ea
te
d
r
u
n
s
.
B
ec
au
s
e
th
e
c
u
r
r
e
n
t
ev
alu
atio
n
r
elies
o
n
r
an
d
o
m
s
em
an
tic
-
lev
el
p
air
in
g
u
n
d
er
a
s
in
g
le
-
r
u
n
s
ettin
g
,
s
o
m
e
d
eg
r
ee
o
f
r
esu
lt
v
ar
iab
ilit
y
m
ay
r
em
ain
.
Fu
tu
r
e
wo
r
k
will
in
v
esti
g
ate
m
o
r
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
S
ema
n
tic
-
a
lig
n
ed
mu
ltimo
d
a
l
h
u
ma
n
a
ctivity
r
ec
o
g
n
itio
n
u
s
in
g
visu
a
l
…
(
Yeeu
n
P
a
r
k
)
2091
co
n
tr
o
lled
s
am
p
le
s
elec
tio
n
an
d
p
air
in
g
s
tr
ateg
ies
af
ter
m
o
d
ality
-
s
p
ec
if
ic
f
in
e
-
tu
n
in
g
,
wh
ich
ar
e
ex
p
ec
ted
to
r
ed
u
ce
v
a
r
iab
ilit
y
an
d
im
p
r
o
v
e
ev
alu
atio
n
s
tab
ilit
y
.
Mo
d
el
p
er
f
o
r
m
an
ce
was
ev
a
lu
ated
u
s
in
g
ac
cu
r
ac
y
an
d
Ma
cr
o
-
F1
s
co
r
e.
Acc
u
r
ac
y
was
u
s
ed
to
s
u
m
m
ar
ize
o
v
er
all
p
r
ed
ictio
n
co
r
r
ec
t
n
ess
,
wh
er
ea
s
Ma
cr
o
-
F1
was
ad
d
itio
n
ally
a
d
o
p
te
d
to
r
ef
lect
class
-
b
alan
ce
d
p
er
f
o
r
m
a
n
ce
u
n
d
e
r
lab
el
im
b
alan
ce
,
wh
ich
is
r
elev
an
t
to
th
e
p
r
esen
t
m
u
ltimo
d
a
l
HAR
s
ettin
g
[
2
2
]
.
Du
r
in
g
ev
al
u
atio
n
,
t
h
e
two
m
o
d
alities
wer
e
alig
n
ed
o
n
ly
at
th
e
s
h
ar
ed
s
em
an
tic
-
g
r
o
u
p
le
v
el
r
ath
er
t
h
an
as
in
s
tan
ce
-
lev
el
s
y
n
ch
r
o
n
ized
v
i
d
eo
-
au
d
io
p
air
s
.
T
h
e
f
in
al
p
r
e
d
ictio
n
was o
b
tain
ed
b
y
co
m
b
i
n
in
g
th
e
im
a
g
e
an
d
au
d
io
lo
g
its
th
r
o
u
g
h
f
ix
ed
s
co
r
e
-
lev
el
late
f
u
s
io
n
.
I
n
th
e
f
in
al
ex
p
e
r
im
en
tal
s
ettin
g
,
a
f
u
s
io
n
weig
h
t
co
n
f
ig
u
r
atio
n
o
f
0
.
6
f
o
r
im
ag
e
an
d
0
.
4
f
o
r
au
d
io
was
ad
o
p
te
d
,
an
d
th
e
im
p
licatio
n
s
o
f
th
is
s
ettin
g
ar
e
f
u
r
th
e
r
d
is
cu
s
s
ed
in
s
ec
tio
n
3
.
3
.
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
3
.
1
.
Unim
o
da
l
m
o
del per
f
o
r
m
a
n
ce
a
na
ly
s
is
T
h
e
im
ag
e
-
b
ased
u
n
im
o
d
al
H
AR
m
o
d
el
ac
h
ie
v
ed
an
a
cc
u
r
ac
y
o
f
8
2
.
0
8
%
an
d
a
Ma
cr
o
-
F1
s
co
r
e
o
f
8
3
.
4
6
%
o
n
th
e
HM
DB
5
1
d
ataset.
T
h
e
d
etailed
p
er
f
o
r
m
a
n
c
e
m
etr
ics
o
f
th
is
im
ag
e
-
b
ased
m
o
d
el
ar
e
v
is
u
ally
s
u
m
m
ar
ized
in
Fig
u
r
e
2
.
As
s
h
o
wn
in
Fig
u
r
e
2
(
a
)
,
r
elativ
el
y
h
ig
h
F1
-
s
co
r
es
wer
e
o
b
s
er
v
ed
in
th
e
o
u
t
d
o
o
r
an
d
in
ter
ac
tio
n
ac
tiv
ity
g
r
o
u
p
s
,
w
h
er
e
d
is
cr
im
in
ativ
e
v
is
u
al
cu
es
ar
e
p
r
o
m
in
en
t.
I
n
co
n
t
r
ast,
th
e
R
esti
n
g
g
r
o
u
p
ex
h
ib
ited
lim
ited
class
if
icatio
n
p
er
f
o
r
m
an
ce
d
u
e
to
its
s
tat
ic
ch
ar
ac
ter
is
tics
an
d
th
e
lack
o
f
d
is
tin
ctiv
e
m
o
tio
n
p
atter
n
s
.
T
h
ese
r
esu
lts
in
d
icate
th
at
f
r
am
e
-
b
ased
v
is
u
al
f
ea
tu
r
es
alo
n
e
ar
e
in
s
u
f
f
icien
t
to
r
eliab
ly
d
is
tin
g
u
is
h
ac
tiv
ities
with
m
in
im
al
v
is
u
al
v
ar
iatio
n
.
Fig
u
r
e
2
(
b
)
s
u
m
m
ar
izes
th
e
o
v
er
all
Acc
u
r
ac
y
an
d
Ma
cr
o
-
F1
s
co
r
e
o
f
th
e
im
ag
e
-
b
ased
u
n
im
o
d
al
m
o
d
el.
(
a)
(
b
)
Fig
u
r
e
2
.
Per
f
o
r
m
an
c
e
o
f
t
h
e
R
esNet
-
1
8
im
ag
e
-
b
ased
u
n
im
o
d
al
m
o
d
el
o
n
t
h
e
HM
DB
5
1
d
ataset
in
(
a)
F1
-
s
co
r
e
b
y
f
iv
e
ac
tiv
ity
g
r
o
u
p
s
an
d
(
b
)
o
v
er
all
ac
c
u
r
ac
y
an
d
Ma
cr
o
-
F1
s
co
r
e
T
h
e
au
d
io
-
b
ased
HAR m
o
d
el
ev
alu
ated
o
n
t
h
e
E
SC
-
5
0
d
atas
et
ac
h
iev
ed
an
Acc
u
r
ac
y
o
f
8
5
.
0
% a
n
d
a
Ma
cr
o
-
F1
s
co
r
e
o
f
8
4
.
0
%.
Fi
g
u
r
e
3
p
r
o
v
i
d
es
a
co
m
p
r
eh
e
n
s
iv
e
o
v
er
v
iew
o
f
th
e
e
v
alu
atio
n
r
esu
lts
f
o
r
th
is
au
d
io
-
b
ased
u
n
im
o
d
al
ap
p
r
o
a
ch
.
As
s
h
o
wn
in
Fig
u
r
e
3
(
a)
,
r
elativ
ely
s
tab
le
F1
-
s
co
r
es
we
r
e
o
b
s
er
v
ed
in
th
e
o
u
td
o
o
r
an
d
in
ter
ac
tio
n
g
r
o
u
p
s
,
wh
er
e
en
v
ir
o
n
m
e
n
tal
s
o
u
n
d
cu
es
p
r
o
v
id
e
m
ea
n
in
g
f
u
l
co
n
t
ex
tu
al
in
f
o
r
m
atio
n
.
Ho
wev
er
,
th
e
o
v
er
all
r
ec
o
g
n
itio
n
p
e
r
f
o
r
m
an
ce
r
em
ain
e
d
lim
ited
b
ec
au
s
e
E
SC
-
5
0
was
o
r
ig
in
ally
d
esig
n
e
d
f
o
r
en
v
ir
o
n
m
en
tal
s
o
u
n
d
class
if
icatio
n
r
ath
er
th
an
d
ir
ec
t
HAR
[
1
7
]
.
T
h
is
r
esu
lt
s
u
g
g
ests
th
at
th
e
au
d
io
m
o
d
ality
ca
n
p
r
o
v
id
e
u
s
ef
u
l
c
o
n
tex
tu
a
l
cu
es
b
u
t
h
as
lim
itatio
n
s
w
h
en
u
s
ed
alo
n
e
f
o
r
co
m
p
r
eh
en
s
iv
e
HAR
task
s
.
Fig
u
r
e
3
(
b
)
p
r
esen
ts
th
e
o
v
er
al
l a
cc
u
r
ac
y
an
d
Ma
cr
o
-
F1
s
co
r
e
o
f
th
e
a
u
d
io
-
b
ased
u
n
im
o
d
al
m
o
d
el.
A
c
t
i
v
i
t
y
G
r
o
u
p
s
F1
-
sc
o
r
e
P
e
r
f
o
r
ma
n
c
e
M
e
t
r
i
c
s
S
c
o
r
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
1
6
,
No
.
4
,
Au
g
u
s
t
20
2
6
:
2
0
8
7
-
2095
2092
(
a)
(
b
)
Fig
u
r
e
3
.
Per
f
o
r
m
an
c
e
o
f
t
h
e
R
esNet
-
1
8
au
d
io
-
b
ased
u
n
im
o
d
al
m
o
d
el
o
n
th
e
E
SC
-
5
0
d
ata
s
et
in
(
a)
F1
-
s
co
r
e
b
y
f
iv
e
ac
tiv
ity
g
r
o
u
p
s
an
d
(
b
)
o
v
er
all
ac
cu
r
a
cy
an
d
Ma
c
r
o
-
F1
s
co
r
e
3
.
2
.
P
er
f
o
r
m
a
nce
co
m
pa
riso
n wit
h la
t
e
f
us
io
n m
o
del
B
y
co
m
b
in
i
n
g
t
h
e
o
u
tp
u
ts
o
f
th
e
im
ag
e
-
b
ased
a
n
d
au
d
io
-
b
as
ed
m
o
d
els
u
s
in
g
th
e
late
f
u
s
io
n
s
tr
ateg
y
,
s
ig
n
if
ican
t
p
er
f
o
r
m
an
ce
im
p
r
o
v
em
en
t
was
ac
h
iev
ed
.
T
h
e
ev
alu
atio
n
was
co
n
d
u
cted
b
y
alig
n
in
g
th
e
im
ag
e
an
d
au
d
io
p
r
ed
ictio
n
s
with
in
th
e
s
am
e
s
h
ar
ed
ac
tiv
ity
g
r
o
u
p
r
ath
er
th
an
u
s
in
g
in
s
tan
ce
-
l
ev
el
s
y
n
ch
r
o
n
ized
v
id
eo
-
au
d
io
p
air
s
.
T
h
e
late
f
u
s
io
n
m
o
d
el
ac
h
iev
ed
an
ac
cu
r
ac
y
o
f
9
2
.
6
7
%
an
d
a
Ma
cr
o
-
F
1
s
co
r
e
o
f
9
2
.
4
9
%,
o
u
tp
er
f
o
r
m
in
g
b
o
th
u
n
im
o
d
al
b
aselin
es.
T
h
e
co
m
p
ar
ativ
e
p
e
r
f
o
r
m
a
n
ce
b
r
ea
k
d
o
wn
o
f
th
e
p
r
o
p
o
s
ed
late
f
u
s
io
n
s
tr
ateg
y
is
illu
s
tr
ated
in
Fig
u
r
e
4
.
(
a)
(
b
)
Fig
u
r
e
4
.
Per
f
o
r
m
an
c
e
co
m
p
ar
is
o
n
o
f
th
e
v
id
eo
–
au
d
io
late
f
u
s
io
n
m
o
d
el
in
(
a)
F1
-
s
co
r
e
b
y
f
iv
e
ac
tiv
ity
g
r
o
u
p
s
an
d
(
b
)
o
v
er
all
ac
cu
r
ac
y
an
d
Ma
cr
o
-
F1
s
co
r
e
A
c
t
i
v
i
t
y
G
r
o
u
p
s
F1
-
sc
o
r
e
P
e
r
f
o
r
ma
n
c
e
M
e
t
r
i
c
s
S
c
o
r
e
F1
-
sc
o
r
e
S
c
o
r
e
A
c
t
i
v
i
t
y
G
r
o
u
p
s
P
e
r
f
o
r
ma
n
c
e
M
e
t
r
i
c
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
S
ema
n
tic
-
a
lig
n
ed
mu
ltimo
d
a
l
h
u
ma
n
a
ctivity
r
ec
o
g
n
itio
n
u
s
in
g
visu
a
l
…
(
Yeeu
n
P
a
r
k
)
2093
As
s
h
o
wn
in
Fig
u
r
e
4
(
a)
,
th
e
f
u
s
io
n
m
o
d
el
ac
h
iev
ed
h
ig
h
e
r
F1
-
s
co
r
es
ac
r
o
s
s
m
o
s
t
ac
tiv
ity
g
r
o
u
p
s
th
an
th
e
u
n
im
o
d
al
m
o
d
els.
T
h
ese
r
esu
lts
in
d
icate
th
at
in
teg
r
atin
g
v
is
u
al
a
n
d
au
d
io
m
o
d
alities
ef
f
ec
tiv
ely
co
m
p
lem
en
ts
th
e
lim
itatio
n
s
o
f
s
in
g
le
-
m
o
d
ality
class
if
icati
o
n
[
1
1
]
,
[
1
5
]
,
[
2
3
]
,
[
2
4
]
.
Fig
u
r
e
4
(
b
)
s
u
m
m
ar
izes
th
e
o
v
e
r
all
Acc
u
r
ac
y
a
n
d
Ma
cr
o
-
F1
s
co
r
e,
s
h
o
win
g
t
h
at
h
ete
r
o
g
en
e
o
u
s
co
n
tex
tu
al
in
f
o
r
m
at
io
n
f
r
o
m
d
if
f
e
r
en
t
m
o
d
alities
ca
n
im
p
r
o
v
e
r
ec
o
g
n
itio
n
r
o
b
u
s
tn
ess
ev
en
u
n
d
er
d
ataset
m
is
m
atch
co
n
d
itio
n
s
.
C
o
m
p
ar
ed
with
p
r
io
r
s
y
n
ch
r
o
n
ized
m
u
ltimo
d
al
HAR
s
tu
d
ies
[
1
1
]
,
[
1
5
]
,
th
e
p
r
esen
t
s
tu
d
y
ev
alu
ates
f
u
s
io
n
u
n
d
e
r
a
s
em
an
tically
alig
n
ed
h
eter
o
g
en
eo
u
s
s
ettin
g
,
h
ig
h
lig
h
tin
g
its
r
o
le
as
a
p
r
ac
tical
late
-
f
u
s
io
n
s
tr
ateg
y
f
o
r
n
o
n
-
s
y
n
ch
r
o
n
ized
m
u
ltimo
d
al
co
n
d
itio
n
s
.
3
.
3
.
F
us
io
n
weig
ht
a
nd
cla
s
s
-
wis
e
perf
o
rm
a
nce
a
na
ly
s
is
A
f
ix
ed
f
u
s
io
n
weig
h
t
co
n
f
ig
u
r
atio
n
o
f
0
.
6
f
o
r
th
e
im
a
g
e
b
r
an
ch
an
d
0
.
4
f
o
r
th
e
au
d
i
o
b
r
an
ch
was
ad
o
p
ted
in
t
h
e
f
in
al
ex
p
er
im
e
n
tal
s
ettin
g
.
Alth
o
u
g
h
an
e
x
p
l
o
r
ato
r
y
weig
h
t
s
ea
r
ch
was
co
n
d
u
cted
d
u
r
in
g
th
e
later
s
tag
e
o
f
th
e
p
r
o
ject,
t
h
e
r
esu
ltin
g
tr
en
d
was
n
o
t
co
n
s
id
er
ed
s
u
f
f
icien
tly
r
o
b
u
s
t
to
s
er
v
e
as
th
e
s
o
le
b
asis
f
o
r
f
i
n
al
weig
h
t
s
elec
tio
n
b
ec
au
s
e
th
e
two
m
o
d
alities
wer
e
alig
n
ed
o
n
ly
at
th
e
s
em
a
n
tic
-
g
r
o
u
p
lev
el
r
ath
er
th
an
as in
s
tan
ce
-
lev
el
s
y
n
ch
r
o
n
ized
p
air
s
.
Acc
o
r
d
in
g
ly
,
th
e
f
in
al
f
u
s
io
n
weig
h
t
was
d
eter
m
in
ed
em
p
i
r
ically
.
A
lar
g
er
weig
h
t
was
a
s
s
ig
n
ed
to
th
e
im
ag
e
b
r
an
ch
b
ec
au
s
e
v
is
u
al
in
f
o
r
m
atio
n
p
r
o
v
id
es
m
o
r
e
d
ir
ec
t
cu
es
f
o
r
HAR
,
wh
er
e
as
au
d
io
in
f
o
r
m
atio
n
m
ain
ly
co
n
tr
ib
u
tes
co
m
p
lem
en
tar
y
co
n
tex
tu
al
in
f
o
r
m
atio
n
.
I
n
ad
d
itio
n
,
th
e
ef
f
ec
tiv
e
au
d
io
s
am
p
le
p
o
o
l
av
ailab
le
af
ter
class
f
ilter
in
g
was
m
o
r
e
lim
ited
t
h
an
t
h
e
ex
tr
ac
ted
im
ag
e
-
f
r
am
e
p
o
o
l,
an
d
f
u
s
io
n
p
air
s
wer
e
co
n
s
tr
u
cted
u
s
in
g
th
e
m
in
im
u
m
n
u
m
b
er
o
f
av
ailab
le
s
am
p
le
s
f
o
r
ea
ch
s
em
an
tic
g
r
o
u
p
.
T
h
i
s
f
u
r
th
er
s
u
p
p
o
r
ted
a
m
o
r
e
co
n
s
er
v
ativ
e
weig
h
tin
g
o
f
th
e
au
d
io
m
o
d
ality
.
T
h
e
late
f
u
s
io
n
m
o
d
el
a
ch
iev
ed
im
p
r
o
v
ed
F1
-
s
co
r
es
ac
r
o
s
s
m
o
s
t
ac
tiv
ity
g
r
o
u
p
s
co
m
p
ar
ed
with
th
e
u
n
im
o
d
al
m
o
d
els,
with
p
ar
ticu
lar
ly
n
o
tab
le
g
ain
s
in
th
e
I
n
ter
ac
tio
n
an
d
Ou
td
o
o
r
g
r
o
u
p
s
.
T
h
ese
f
in
d
in
g
s
s
u
g
g
est
th
at
v
is
u
al
in
f
o
r
m
ati
o
n
p
lay
s
a
d
o
m
in
an
t
r
o
le
in
o
v
er
all
r
ec
o
g
n
itio
n
,
w
h
ile
au
d
io
in
f
o
r
m
atio
n
ca
n
p
r
o
v
id
e
c
o
m
p
lem
en
ta
r
y
co
n
t
ex
tu
al
cu
es
in
co
n
tex
t
-
d
ep
e
n
d
en
t
ac
tiv
ities
[
2
4
]
,
[
2
5
]
.
C
o
n
s
is
ten
t
with
p
r
io
r
au
d
io
-
v
is
u
al
f
u
s
io
n
s
tu
d
ies
[
1
5
]
,
t
h
e
p
r
o
p
o
s
ed
f
u
s
io
n
s
h
o
we
d
p
r
ac
tical
b
en
e
f
its
u
n
d
er
a
s
h
ar
ed
s
em
an
tic
lab
el
s
p
ac
e.
T
h
er
ef
o
r
e,
th
e
0
.
6
:0
.
4
r
atio
s
h
o
u
ld
b
e
in
ter
p
r
ete
d
as
a
n
em
p
ir
ically
ch
o
s
en
p
r
ac
tical
s
ettin
g
r
ath
er
th
an
a
r
ig
o
r
o
u
s
ly
o
p
tim
ized
f
u
s
io
n
r
atio
.
4.
CO
NCLU
SI
O
N
T
h
is
p
ap
er
p
r
esen
ted
a
m
u
lti
m
o
d
al
HAR
f
r
am
ewo
r
k
th
at
in
teg
r
ates
v
is
u
al
an
d
au
d
io
i
n
f
o
r
m
atio
n
u
s
in
g
a
late
f
u
s
io
n
s
tr
ateg
y
.
Fo
r
th
e
v
is
u
al
m
o
d
ality
,
a
R
esNet
-
1
8
m
o
d
el
was tr
ain
ed
u
s
in
g
f
r
am
e
-
b
ased
in
p
u
ts
s
am
p
led
f
r
o
m
th
e
HM
DB
5
1
d
ataset.
Fo
r
th
e
au
d
io
m
o
d
ality
,
en
v
ir
o
n
m
en
tal
s
o
u
n
d
s
am
p
le
s
f
r
o
m
th
e
E
SC
-
50
d
ataset
wer
e
co
n
v
er
ted
in
to
6
4
-
b
in
Me
l
-
s
p
ec
tr
o
g
r
am
r
ep
r
esen
tatio
n
s
an
d
p
r
o
ce
s
s
ed
u
s
in
g
th
e
s
am
e
n
etwo
r
k
ar
ch
itectu
r
e.
Du
r
in
g
in
f
er
en
ce
,
th
e
lo
g
its
o
f
th
e
two
u
n
im
o
d
a
l m
o
d
els we
r
e
co
m
b
in
ed
th
r
o
u
g
h
weig
h
ted
lin
ea
r
f
u
s
io
n
to
p
r
o
d
u
ce
th
e
f
in
al
ac
tiv
ity
p
r
ed
ictio
n
.
E
x
p
er
im
en
tal
r
esu
lts
s
h
o
wed
th
at
th
e
p
r
o
p
o
s
ed
late
f
u
s
io
n
m
o
d
el
o
u
tp
er
f
o
r
m
e
d
th
e
u
n
im
o
d
al
b
aselin
es,
ac
h
iev
in
g
an
Acc
u
r
ac
y
o
f
9
2
.
6
7
%
an
d
a
Ma
cr
o
-
F1
s
co
r
e
o
f
9
2
.
4
9
%.
No
tab
le
p
er
f
o
r
m
an
ce
im
p
r
o
v
em
e
n
ts
wer
e
o
b
s
er
v
ed
in
th
e
Ou
td
o
o
r
,
I
n
ter
ac
tio
n
,
an
d
Activ
e
ac
tiv
ity
g
r
o
u
p
s
.
T
h
ese
f
in
d
in
g
s
s
u
g
g
est
th
at
th
e
au
d
io
m
o
d
ality
ca
n
p
r
o
v
id
e
co
m
p
le
m
en
tar
y
en
v
ir
o
n
m
en
tal
co
n
tex
t
in
v
is
u
ally
am
b
ig
u
o
u
s
o
r
c
o
n
tex
t
-
d
ep
en
d
e
n
t
ac
tiv
ity
s
ce
n
ar
io
s
.
Fo
r
r
elativ
ely
s
tatic
ac
tiv
ity
g
r
o
u
p
s
s
u
ch
as
R
esti
n
g
,
th
e
d
ir
ec
t
co
n
tr
ib
u
tio
n
o
f
au
d
io
in
f
o
r
m
atio
n
was m
o
r
e
li
m
ited
.
T
h
is
s
tu
d
y
ad
d
r
ess
ed
d
ataset
h
eter
o
g
en
eity
b
y
r
eo
r
g
a
n
izin
g
c
lass
es
f
r
o
m
HM
DB
5
1
an
d
E
SC
-
5
0
in
to
s
h
ar
ed
s
em
an
tic
ac
tiv
ity
g
r
o
u
p
s
,
th
er
eb
y
en
a
b
lin
g
s
co
r
e
-
le
v
el
m
u
ltimo
d
al
f
u
s
io
n
u
n
d
er
m
is
m
atch
ed
d
ataset
co
n
d
itio
n
s
.
T
h
e
r
esu
lts
s
u
g
g
es
t
th
at
a
late
f
u
s
io
n
-
b
ased
ap
p
r
o
ac
h
ca
n
s
er
v
e
as
a
p
r
ac
tical
ex
p
lo
r
ato
r
y
b
aselin
e
f
o
r
m
u
ltim
o
d
al
HAR in
s
ettin
g
s
wh
er
e
h
eter
o
g
en
eo
u
s
d
ata
s
o
u
r
ce
s
m
u
s
t b
e
jo
i
n
tly
u
tili
ze
d
.
Desp
ite
th
ese
ad
v
an
tag
es,
th
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
h
as
s
ev
er
al
lim
itatio
n
s
.
B
ec
au
s
e
th
e
m
o
d
el
r
elies
o
n
s
co
r
e
-
le
v
el
late
f
u
s
io
n
,
it
d
o
es
n
o
t
e
x
p
licitly
ca
p
tu
r
e
f
i
n
e
-
g
r
ain
e
d
cr
o
s
s
-
m
o
d
al
in
ter
a
ctio
n
s
at
th
e
f
ea
tu
r
e
lev
el.
I
n
ad
d
itio
n
,
th
e
v
is
u
al
an
d
au
d
io
d
ata
wer
e
n
o
t
s
y
n
ch
r
o
n
ized
at
th
e
in
s
tan
ce
lev
el
b
u
t
wer
e
alig
n
ed
th
r
o
u
g
h
s
h
ar
ed
s
em
an
tic
ac
tiv
ity
g
r
o
u
p
s
.
T
h
e
r
ef
o
r
e
,
th
e
cu
r
r
en
t
ev
alu
atio
n
s
h
o
u
ld
b
e
in
te
r
p
r
eted
as
s
em
an
tic
-
lev
el
m
u
ltimo
d
al
f
u
s
io
n
u
n
d
er
d
ataset
m
is
m
atch
co
n
d
itio
n
s
r
ath
er
th
an
f
u
lly
s
y
n
ch
r
o
n
ized
m
u
ltimo
d
a
l
r
ec
o
g
n
itio
n
.
Mo
r
eo
v
er
,
b
ec
a
u
s
e
th
e
p
r
esen
t
ev
alu
atio
n
was
co
n
d
u
cted
u
n
d
er
a
s
in
g
le
f
ix
ed
s
ettin
g
with
r
an
d
o
m
s
em
an
tic
-
lev
el
p
air
in
g
,
s
o
m
e
v
ar
ia
b
ilit
y
m
ay
r
em
a
in
.
Fu
tu
r
e
wo
r
k
will
f
o
cu
s
o
n
ev
alu
ati
o
n
s
u
s
in
g
s
y
n
ch
r
o
n
ize
d
m
u
ltimo
d
al
d
a
tasets
,
m
o
r
e
co
n
tr
o
lled
s
am
p
le
s
elec
tio
n
an
d
p
air
in
g
s
t
r
ateg
ies,
an
d
m
o
r
e
ad
v
an
ce
d
f
u
s
io
n
o
r
s
p
atio
tem
p
o
r
al
m
o
d
elin
g
ap
p
r
o
ac
h
es.
T
h
e
p
r
o
p
o
s
ed
f
r
am
ewo
r
k
m
ay
also
p
r
o
v
id
e
a
u
s
ef
u
l
b
asis
f
o
r
f
u
tu
r
e
ex
ten
s
io
n
s
to
wea
r
ab
le
HAR en
v
ir
o
n
m
e
n
ts
,
in
clu
d
in
g
s
m
ar
t
g
lass
es.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
1
6
,
No
.
4
,
Au
g
u
s
t
20
2
6
:
2
0
8
7
-
2095
2094
ACK
NO
WL
E
DG
M
E
N
T
S
T
h
is
wo
r
k
was
s
u
p
p
o
r
te
d
b
y
t
h
e
Natio
n
al
R
esear
ch
Fo
u
n
d
at
io
n
o
f
Ko
r
ea
(
NR
F)
g
r
a
n
t
f
u
n
d
ed
b
y
th
e
Ko
r
ea
g
o
v
er
n
m
e
n
t
(
MSI
T
)
(
N
o
.
R
S
-
2025
-
1
9
9
2
2
9
6
8
)
.
AUTHO
R
CO
NT
RI
B
UT
I
O
NS ST
A
T
E
M
E
N
T
T
h
is
jo
u
r
n
al
u
s
es
th
e
C
o
n
tr
ib
u
to
r
R
o
les
T
ax
o
n
o
m
y
(
C
R
ed
iT)
to
r
ec
o
g
n
ize
in
d
iv
id
u
al
au
th
o
r
co
n
tr
ib
u
tio
n
s
,
r
ed
u
ce
au
th
o
r
s
h
ip
d
is
p
u
tes,
an
d
f
ac
ilit
ate
co
llab
o
r
atio
n
.
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
Yee
u
n
Par
k
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
J
o
o
n
h
o
B
y
u
n
✓
✓
✓
✓
Siwo
o
B
y
u
n
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
C
:
C
o
n
c
e
p
t
u
a
l
i
z
a
t
i
o
n
M
:
M
e
t
h
o
d
o
l
o
g
y
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
d
a
t
i
o
n
Fo
:
Fo
r
mal
a
n
a
l
y
s
i
s
I
:
I
n
v
e
s
t
i
g
a
t
i
o
n
R
:
R
e
so
u
r
c
e
s
D
:
D
a
t
a
C
u
r
a
t
i
o
n
O
:
W
r
i
t
i
n
g
-
O
r
i
g
i
n
a
l
D
r
a
f
t
E
:
W
r
i
t
i
n
g
-
R
e
v
i
e
w
&
E
d
i
t
i
n
g
Vi
:
Vi
su
a
l
i
z
a
t
i
o
n
Su
:
Su
p
e
r
v
i
s
i
o
n
P
:
P
r
o
j
e
c
t
a
d
mi
n
i
st
r
a
t
i
o
n
Fu
:
Fu
n
d
i
n
g
a
c
q
u
i
si
t
i
o
n
CO
NF
L
I
C
T
O
F
I
N
T
E
R
E
S
T
ST
A
T
E
M
E
NT
T
h
e
au
th
o
r
s
d
ec
lar
e
n
o
co
n
f
lict o
f
in
ter
est.
DATA AV
AI
L
AB
I
L
I
T
Y
T
h
is
s
tu
d
y
u
s
ed
p
u
b
licly
av
ail
ab
le
d
atasets
,
HM
D
B
5
1
an
d
E
SC
-
5
0
,
wh
ich
ca
n
b
e
ac
ce
s
s
ed
th
r
o
u
g
h
th
eir
o
f
f
icial
r
ep
o
s
ito
r
ies.
No
n
ew
d
ataset
was
cr
ea
ted
in
t
h
is
s
tu
d
y
.
T
h
e
im
p
lem
en
tatio
n
co
d
e
r
elate
d
to
th
e
m
u
ltimo
d
al
HAR f
r
am
ewo
r
k
ca
n
b
e
m
a
d
e
av
ailab
le
f
r
o
m
th
e
co
r
r
esp
o
n
d
in
g
au
th
o
r
u
p
o
n
r
ea
s
o
n
ab
le
r
eq
u
est.
RE
F
E
R
E
NC
E
S
[
1
]
F
.
G
u
,
M
.
H
.
C
h
u
n
g
,
M
.
C
h
i
g
n
e
l
l
,
S
.
V
a
l
a
e
e
,
B
.
Z
h
o
u
,
a
n
d
X
.
Li
u
,
“
A
s
u
r
v
e
y
o
n
d
e
e
p
l
e
a
r
n
i
n
g
f
o
r
h
u
ma
n
a
c
t
i
v
i
t
y
r
e
c
o
g
n
i
t
i
o
n
,
”
AC
M
C
o
m
p
u
t
i
n
g
S
u
rv
e
y
s
,
v
o
l
.
5
4
,
n
o
.
8
,
p
p
.
1
–
3
4
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
4
5
/
3
4
7
2
2
9
0
.
[
2
]
J.
S
h
i
n
,
N
.
H
a
ssa
n
,
A
.
S
.
M
.
M
i
a
h
,
a
n
d
S
.
N
i
s
h
i
m
u
r
a
,
“
A
c
o
m
p
r
e
h
e
n
s
i
v
e
met
h
o
d
o
l
o
g
i
c
a
l
s
u
r
v
e
y
o
f
h
u
m
a
n
a
c
t
i
v
i
t
y
r
e
c
o
g
n
i
t
i
o
n
a
c
r
o
ss
d
i
v
e
r
se
d
a
t
a
m
o
d
a
l
i
t
i
e
s,
”
S
e
n
s
o
rs
,
v
o
l
.
2
5
,
n
o
.
1
3
,
p
.
4
0
2
8
,
2
0
2
5
,
d
o
i
:
1
0
.
3
3
9
0
/
s
2
5
1
3
4
0
2
8
.
[
3
]
A
.
B
u
l
l
i
n
g
,
U
.
B
l
a
n
k
e
,
a
n
d
B
.
S
c
h
i
e
l
e
,
“
A
t
u
t
o
r
i
a
l
o
n
h
u
m
a
n
a
c
t
i
v
i
t
y
r
e
c
o
g
n
i
t
i
o
n
u
s
i
n
g
b
o
d
y
-
w
o
r
n
i
n
e
r
t
i
a
l
se
n
s
o
r
s,”
A
C
M
C
o
m
p
u
t
i
n
g
S
u
r
v
e
y
s
,
v
o
l
.
4
6
,
n
o
.
3
,
p
.
3
3
,
2
0
1
4
,
d
o
i
:
1
0
.
1
1
4
5
/
2
4
9
9
6
2
1
.
[
4
]
K
.
S
i
mo
n
y
a
n
a
n
d
A
.
Z
i
sserm
a
n
,
“
T
w
o
-
st
r
e
a
m
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
t
w
o
r
k
s
f
o
r
a
c
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
i
n
v
i
d
e
o
s,
”
i
n
A
d
v
a
n
c
e
s
i
n
N
e
u
ra
l
I
n
f
o
rm
a
t
i
o
n
Pr
o
c
e
ssi
n
g
S
y
st
e
m
s
,
2
0
1
4
,
v
o
l
.
1
,
n
o
.
J
a
n
u
a
r
y
,
p
p
.
5
6
8
–
5
7
6
.
[
5
]
J.
C
a
r
r
e
i
r
a
a
n
d
A
.
Zi
ss
e
r
ma
n
,
“
Q
u
o
V
a
d
i
s,
a
c
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
?
A
n
e
w
m
o
d
e
l
a
n
d
t
h
e
k
i
n
e
t
i
c
s
d
a
t
a
se
t
,
”
i
n
Pro
c
e
e
d
i
n
g
s
-
3
0
t
h
I
EEE
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
Pa
t
t
e
r
n
Re
c
o
g
n
i
t
i
o
n
,
2
0
1
7
,
v
o
l
.
2
0
1
7
,
p
p
.
4
7
2
4
–
4
7
3
3
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
7
.
5
0
2
.
[
6
]
S
.
H
e
r
s
h
e
y
e
t
a
l
.
,
“
C
N
N
a
r
c
h
i
t
e
c
t
u
r
e
s
f
o
r
l
a
r
g
e
-
sc
a
l
e
a
u
d
i
o
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
i
n
I
C
A
S
S
P,
I
EE
E
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
Ac
o
u
s
t
i
c
s
,
S
p
e
e
c
h
a
n
d
S
i
g
n
a
l
Pr
o
c
e
ss
i
n
g
-
Pro
c
e
e
d
i
n
g
s
,
2
0
1
7
,
p
p
.
1
3
1
–
1
3
5
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
A
S
S
P
.
2
0
1
7
.
7
9
5
2
1
3
2
.
[
7
]
J.
S
a
l
a
m
o
n
a
n
d
J.
P
.
B
e
l
l
o
,
“
D
e
e
p
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s
a
n
d
d
a
t
a
a
u
g
me
n
t
a
t
i
o
n
f
o
r
e
n
v
i
r
o
n
m
e
n
t
a
l
s
o
u
n
d
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
I
EEE
S
i
g
n
a
l
Pr
o
c
e
ssi
n
g
L
e
t
t
e
rs
,
v
o
l
.
2
4
,
n
o
.
3
,
p
p
.
2
7
9
–
2
8
3
,
2
0
1
7
,
d
o
i
:
1
0
.
1
1
0
9
/
LSP
.
2
0
1
7
.
2
6
5
7
3
8
1
.
[
8
]
T.
H
e
i
t
t
o
l
a
,
A
.
M
e
s
a
r
o
s,
A
.
Er
o
n
e
n
,
a
n
d
T.
V
i
r
t
a
n
e
n
,
“
C
o
n
t
e
x
t
-
d
e
p
e
n
d
e
n
t
s
o
u
n
d
e
v
e
n
t
d
e
t
e
c
t
i
o
n
,
”
Eu
r
a
si
p
J
o
u
r
n
a
l
o
n
Au
d
i
o
,
S
p
e
e
c
h
,
a
n
d
Mu
s
i
c
Pr
o
c
e
s
si
n
g
,
v
o
l
.
2
0
1
3
,
n
o
.
1
,
p
.
1
,
2
0
1
3
,
d
o
i
:
1
0
.
1
1
8
6
/
1
6
8
7
-
4
7
2
2
-
2
0
1
3
-
1.
[
9
]
S
.
K
.
Y
a
d
a
v
,
K
.
Ti
w
a
r
i
,
H
.
M
.
P
a
n
d
e
y
,
a
n
d
S
.
A
.
A
k
b
a
r
,
“
A
r
e
v
i
e
w
o
f
mu
l
t
i
mo
d
a
l
h
u
ma
n
a
c
t
i
v
i
t
y
r
e
c
o
g
n
i
t
i
o
n
w
i
t
h
s
p
e
c
i
a
l
e
mp
h
a
s
i
s
o
n
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
a
p
p
l
i
c
a
t
i
o
n
s,
c
h
a
l
l
e
n
g
e
s
a
n
d
f
u
t
u
r
e
d
i
r
e
c
t
i
o
n
s,”
K
n
o
w
l
e
d
g
e
-
B
a
se
d
S
y
st
e
m
s
,
v
o
l
.
2
2
3
,
p
.
1
0
6
9
7
0
,
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
k
n
o
s
y
s.
2
0
2
1
.
1
0
6
9
7
0
.
[
1
0
]
T.
B
a
l
t
r
u
sai
t
i
s
,
C
.
A
h
u
j
a
,
a
n
d
L.
P
.
M
o
r
e
n
c
y
,
“
M
u
l
t
i
mo
d
a
l
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
:
A
s
u
r
v
e
y
a
n
d
t
a
x
o
n
o
m
y
,
”
I
E
E
E
T
ra
n
s
a
c
t
i
o
n
s
o
n
Pa
t
t
e
r
n
An
a
l
y
si
s
a
n
d
Ma
c
h
i
n
e
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
4
1
,
n
o
.
2
,
p
p
.
4
2
3
–
4
4
3
,
2
0
1
9
,
d
o
i
:
1
0
.
1
1
0
9
/
TPA
M
I
.
2
0
1
8
.
2
7
9
8
6
0
7
.
[
1
1
]
Q
.
W
u
,
Z
.
W
a
n
g
,
F
.
D
e
n
g
,
a
n
d
D
.
D
.
F
e
n
g
,
“
R
e
a
l
i
st
i
c
h
u
m
a
n
a
c
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
w
i
t
h
a
u
d
i
o
c
o
n
t
e
x
t
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
D
i
g
i
t
a
l
I
m
a
g
e
C
o
m
p
u
t
i
n
g
:
T
e
c
h
n
i
q
u
e
s
a
n
d
A
p
p
l
i
c
a
t
i
o
n
s
(
D
I
C
T
A)
,
2
0
1
0
,
p
p
.
2
8
8
–
2
9
3
,
d
o
i
:
1
0
.
1
1
0
9
/
D
I
C
TA
.
2
0
1
0
.
5
7
.
[
1
2
]
A
.
S
h
a
b
a
n
a
n
d
S
.
Y
o
u
sef
i
,
“
M
u
l
t
i
mo
d
a
l
d
e
e
p
l
e
a
r
n
i
n
g
,
”
i
n
S
p
r
i
n
g
e
r
O
p
t
i
m
i
z
a
t
i
o
n
a
n
d
I
t
s
A
p
p
l
i
c
a
t
i
o
n
s
,
2
0
2
4
,
v
o
l
.
2
1
1
,
p
p
.
2
0
9
–
2
1
9
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
0
3
1
-
5
3
0
9
2
-
0
_
1
0
.
[
1
3
]
T.
Ji
a
o
,
C
.
G
u
o
,
X
.
F
e
n
g
,
Y
.
C
h
e
n
,
a
n
d
J
.
S
o
n
g
,
“
A
c
o
m
p
r
e
h
e
n
si
v
e
su
r
v
e
y
o
n
d
e
e
p
l
e
a
r
n
i
n
g
mu
l
t
i
-
mo
d
a
l
f
u
si
o
n
:
M
e
t
h
o
d
s
,
t
e
c
h
n
o
l
o
g
i
e
s
a
n
d
a
p
p
l
i
c
a
t
i
o
n
s,
”
C
o
m
p
u
t
e
rs
,
Ma
t
e
r
i
a
l
s
a
n
d
C
o
n
t
i
n
u
a
,
v
o
l
.
8
0
,
n
o
.
1
,
p
p
.
1
–
3
5
,
2
0
2
4
,
d
o
i
:
1
0
.
3
2
6
0
4
/
c
mc
.
2
0
2
4
.
0
5
3
2
0
4
.
[
1
4
]
N
.
N
e
v
e
r
o
v
a
,
C
.
W
o
l
f
,
G
.
Ta
y
l
o
r
,
a
n
d
F
.
N
e
b
o
u
t
,
“
M
o
d
D
r
o
p
:
A
d
a
p
t
i
v
e
m
u
l
t
i
-
m
o
d
a
l
g
e
s
t
u
r
e
r
e
c
o
g
n
i
t
i
o
n
,
”
I
EE
E
T
ra
n
sa
c
t
i
o
n
s
o
n
Pa
t
t
e
r
n
An
a
l
y
si
s
a
n
d
Ma
c
h
i
n
e
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
3
8
,
n
o
.
8
,
p
p
.
1
6
9
2
–
1
7
0
6
,
2
0
1
6
,
d
o
i
:
1
0
.
1
1
0
9
/
TPA
M
I
.
2
0
1
5
.
2
4
6
1
5
4
4
.
[
1
5
]
E.
K
a
z
a
k
o
s
,
A
.
N
a
g
r
a
n
i
,
A
.
Z
i
sser
man
,
a
n
d
D
i
.
D
a
me
n
,
“
EPI
C
-
f
u
si
o
n
:
A
u
d
i
o
-
v
i
s
u
a
l
t
e
m
p
o
r
a
l
b
i
n
d
i
n
g
f
o
r
e
g
o
c
e
n
t
r
i
c
a
c
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
,
”
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
I
EEE
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
,
p
p
.
5
4
9
1
–
5
5
0
0
,
2
0
1
9
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
C
V
.
2
0
1
9
.
0
0
5
5
9
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
S
ema
n
tic
-
a
lig
n
ed
mu
ltimo
d
a
l
h
u
ma
n
a
ctivity
r
ec
o
g
n
itio
n
u
s
in
g
visu
a
l
…
(
Yeeu
n
P
a
r
k
)
2095
[
1
6
]
H
.
K
u
e
h
n
e
,
H
.
J
h
u
a
n
g
,
E.
G
a
r
r
o
t
e
,
T
.
P
o
g
g
i
o
,
a
n
d
T.
S
e
r
r
e
,
“
H
M
D
B
:
A
l
a
r
g
e
v
i
d
e
o
d
a
t
a
b
a
s
e
f
o
r
h
u
ma
n
m
o
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
I
E
EE
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r V
i
si
o
n
,
2
0
1
1
,
p
p
.
2
5
5
6
–
2
5
6
3
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
C
V
.
2
0
1
1
.
6
1
2
6
5
4
3
.
[
1
7
]
K
.
J.
P
i
c
z
a
k
,
“
ESC
:
D
a
t
a
s
e
t
f
o
r
e
n
v
i
r
o
n
m
e
n
t
a
l
s
o
u
n
d
c
l
a
s
si
f
i
c
a
t
i
o
n
,
”
i
n
MM
2
0
1
5
-
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
2
0
1
5
AC
M
M
u
l
t
i
m
e
d
i
a
C
o
n
f
e
re
n
c
e
,
2
0
1
5
,
p
p
.
1
0
1
5
–
1
0
1
8
,
d
o
i
:
1
0
.
1
1
4
5
/
2
7
3
3
3
7
3
.
2
8
0
6
3
9
0
.
[
1
8
]
D
.
S
.
P
.
e
t
a
l
.
,
“
S
p
e
c
A
u
g
me
n
t
:
A
si
m
p
l
e
d
a
t
a
a
u
g
m
e
n
t
a
t
i
o
n
m
e
t
h
o
d
f
o
r
a
u
t
o
ma
t
i
c
sp
e
e
c
h
r
e
c
o
g
n
i
t
i
o
n
,
”
i
n
Pr
o
c
.
I
n
t
e
rs
p
e
e
c
h
2
0
1
9
,
2
0
1
9
,
p
p
.
2
6
1
3
–
2
6
1
7
,
d
o
i
:
1
0
.
2
1
4
3
7
/
I
n
t
e
r
sp
e
e
c
h
.
2
0
1
9
-
2
6
8
0
.
[
1
9
]
K
.
H
e
,
X
.
Z
h
a
n
g
,
S
.
R
e
n
,
a
n
d
J
.
S
u
n
,
“
D
e
e
p
r
e
s
i
d
u
a
l
l
e
a
r
n
i
n
g
f
o
r
i
ma
g
e
r
e
c
o
g
n
i
t
i
o
n
,
”
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
I
EEE
C
o
m
p
u
t
e
r
S
o
c
i
e
t
y
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
Pa
t
t
e
r
n
Re
c
o
g
n
i
t
i
o
n
,
v
o
l
.
2
0
1
6
-
D
e
c
e
mb
e
r
,
p
p
.
7
7
0
–
7
7
8
,
2
0
1
6
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
6
.
9
0
.
[
2
0
]
L.
W
a
n
g
e
t
a
l
.
,
“
Te
m
p
o
r
a
l
se
g
m
e
n
t
n
e
t
w
o
r
k
s:
T
o
w
a
r
d
s g
o
o
d
p
r
a
c
t
i
c
e
s
f
o
r
d
e
e
p
a
c
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
,
”
i
n
L
e
c
t
u
r
e
N
o
t
e
s
i
n
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
(
i
n
c
l
u
d
i
n
g
s
u
b
ser
i
e
s
L
e
c
t
u
r
e
N
o
t
e
s
i
n
Ar
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
L
e
c
t
u
re
N
o
t
e
s i
n
Bi
o
i
n
f
o
rm
a
t
i
c
s)
,
2
0
1
6
,
v
o
l
.
9
9
1
2
LN
C
S
,
p
p
.
2
0
–
3
6
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
319
-
4
6
4
8
4
-
8
_
2
.
[
2
1
]
M
.
Z
o
l
f
a
g
h
a
r
i
,
K
.
S
i
n
g
h
,
a
n
d
T.
B
r
o
x
,
“
EC
O
:
Ef
f
i
c
i
e
n
t
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
t
w
o
r
k
f
o
r
o
n
l
i
n
e
v
i
d
e
o
u
n
d
e
r
st
a
n
d
i
n
g
,
”
i
n
L
e
c
t
u
re
N
o
t
e
s
i
n
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
(
i
n
c
l
u
d
i
n
g
s
u
b
s
e
ri
e
s
L
e
c
t
u
re
N
o
t
e
s
i
n
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
L
e
c
t
u
re
N
o
t
e
s
i
n
Bi
o
i
n
f
o
rm
a
t
i
c
s)
,
2
0
1
8
,
v
o
l
.
1
1
2
0
6
LN
C
S
,
p
p
.
7
1
3
–
7
3
0
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
0
3
0
-
0
1
2
1
6
-
8
_
4
3
.
[
2
2
]
M
.
S
o
k
o
l
o
v
a
a
n
d
G
.
L
a
p
a
l
me,
“
A
sy
st
e
ma
t
i
c
a
n
a
l
y
s
i
s
o
f
p
e
r
f
o
r
m
a
n
c
e
me
a
su
r
e
s
f
o
r
c
l
a
ssi
f
i
c
a
t
i
o
n
t
a
s
k
s,”
I
n
f
o
rm
a
t
i
o
n
Pro
c
e
ss
i
n
g
a
n
d
Ma
n
a
g
e
m
e
n
t
,
v
o
l
.
4
5
,
n
o
.
4
,
p
p
.
4
2
7
–
4
3
7
,
2
0
0
9
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
p
m
.
2
0
0
9
.
0
3
.
0
0
2
.
[
2
3
]
Y
.
Zh
a
n
g
,
H
.
D
o
u
g
h
t
y
,
L.
S
h
a
o
,
a
n
d
C
.
G
.
M
.
S
n
o
e
k
,
“
A
u
d
i
o
-
a
d
a
p
t
i
v
e
a
c
t
i
v
i
t
y
r
e
c
o
g
n
i
t
i
o
n
a
c
r
o
ss v
i
d
e
o
d
o
ma
i
n
s
,
”
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
I
EE
E
C
o
m
p
u
t
e
r
S
o
c
i
e
t
y
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
V
i
si
o
n
a
n
d
P
a
t
t
e
r
n
Re
c
o
g
n
i
t
i
o
n
,
v
o
l
.
2
0
2
2
-
J
u
n
e
,
p
p
.
1
3
7
8
1
–
1
3
7
9
0
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
5
2
6
8
8
.
2
0
2
2
.
0
1
3
4
2
.
[
2
4
]
R
.
G
a
o
,
T.
H
.
O
h
,
K
.
G
r
a
u
ma
n
,
a
n
d
L
.
T
o
r
r
e
sa
n
i
,
“
Li
s
t
e
n
t
o
l
o
o
k
:
A
c
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
b
y
p
r
e
v
i
e
w
i
n
g
a
u
d
i
o
,
”
i
n
P
r
o
c
e
e
d
i
n
g
s
o
f
t
h
e
I
EEE
C
o
m
p
u
t
e
r
S
o
c
i
e
t
y
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
P
a
t
t
e
r
n
Re
c
o
g
n
i
t
i
o
n
,
2
0
2
0
,
p
p
.
1
0
4
5
4
–
1
0
4
6
4
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
4
2
6
0
0
.
2
0
2
0
.
0
1
0
4
7
.
[
2
5
]
M
.
B
.
S
h
a
i
k
h
,
D
.
C
h
a
i
,
S
.
M
.
S
.
I
sl
a
m
,
a
n
d
N
.
A
k
h
t
a
r
,
“
M
u
l
t
i
m
o
d
a
l
f
u
si
o
n
f
o
r
a
u
d
i
o
-
i
ma
g
e
a
n
d
v
i
d
e
o
a
c
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
,
”
N
e
u
r
a
l
C
o
m
p
u
t
i
n
g
a
n
d
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
3
6
,
n
o
.
1
0
,
p
p
.
5
4
9
9
–
5
5
1
3
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
0
7
/
s
0
0
5
2
1
-
0
2
3
-
0
9
1
8
6
-
5.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
Ye
e
u
n
P
a
r
k
r
e
c
e
iv
e
d
h
e
r
B.
S
.
d
e
g
re
e
in
in
f
o
rm
a
ti
o
n
,
e
lec
tri
c
a
l,
a
n
d
e
lec
tro
n
ic
e
n
g
in
e
e
rin
g
fr
o
m
An
y
a
n
g
U
n
iv
e
r
sity
,
An
y
a
n
g
,
S
o
u
th
K
o
re
a
.
S
h
e
wa
s
a
re
c
ip
ien
t
o
f
t
h
e
ARI
Aw
a
rd
a
t
th
e
An
y
a
n
g
Un
iv
e
rsit
y
Ca
p
sto
n
e
De
sig
n
Co
m
p
e
ti
ti
o
n
fo
r
th
e
AI
-
b
a
se
d
e
m
o
ti
o
n
-
a
wa
re
p
e
rso
n
a
li
z
e
d
se
rv
ice
p
ro
je
c
t.
He
r
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
in
telli
g
e
n
t
we
b
a
n
d
m
o
b
il
e
a
p
p
li
c
a
ti
o
n
s,
m
a
c
h
in
e
lea
rn
i
n
g
,
d
e
e
p
lea
rn
in
g
,
m
u
lt
imo
d
a
l
le
a
rn
in
g
,
a
n
d
c
o
n
tex
t
-
a
wa
re
p
e
rso
n
a
li
z
e
d
se
rv
ice
s.
S
h
e
is
p
a
rti
c
u
larly
f
o
c
u
se
d
o
n
th
e
p
ra
c
ti
c
a
l
imp
lem
e
n
tatio
n
o
f
u
se
r
-
c
e
n
tere
d
in
tell
ig
e
n
t
sy
ste
m
s
i
n
m
o
b
il
e
a
n
d
Io
T
e
n
v
ir
o
n
m
e
n
ts.
In
t
h
is
st
u
d
y
,
s
h
e
c
o
n
tri
b
u
ted
t
o
e
x
p
e
rime
n
tal
d
e
sig
n
,
fra
m
e
wo
rk
imp
lem
e
n
tatio
n
,
re
su
lt
a
n
a
ly
sis,
a
n
d
m
a
n
u
sc
rip
t
writi
n
g
.
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
stu
d
y
e
2
0
@g
m
a
il
.
c
o
m
.
J
o
o
n
h
o
By
u
n
re
c
e
iv
e
d
h
is
B.
S
.
d
e
g
re
e
in
e
lec
tro
n
ic
e
n
g
in
e
e
ri
n
g
fr
o
m
Ko
o
k
m
in
Un
iv
e
rsity
,
S
e
o
u
l,
S
o
u
t
h
K
o
re
a
,
in
2
0
2
2
.
F
r
o
m
M
a
rc
h
2
0
2
3
to
Ja
n
u
a
ry
2
0
2
4
,
h
e
wa
s
a
n
AI
re
se
a
rc
h
e
r
with
th
e
In
d
u
strial
A
Io
T
re
se
a
rc
h
c
e
n
ter
a
t
An
y
a
n
g
Un
iv
e
rsity
,
An
y
a
n
g
,
S
o
u
th
Ko
re
a
.
He
is
in
a
n
M
.
S
.
a
t
th
e
G
ra
d
u
a
te
S
c
h
o
o
l
o
f
Artifi
c
ial
In
tell
i
g
e
n
c
e
a
t
S
o
g
a
n
g
Un
iv
e
rsity
,
S
e
o
u
l,
S
o
u
th
Ko
re
a
.
He
is a rec
ip
ien
t
o
f
b
e
st p
a
p
e
r
a
wa
rd
in
2
0
2
3
a
n
d
2
0
2
5
.
His m
a
in
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
c
o
m
p
u
ter
v
isi
o
n
,
Io
T,
m
a
c
h
in
e
lea
rn
i
n
g
,
d
e
e
p
l
e
a
rn
in
g
,
sm
a
rt
g
las
s,
v
i
d
e
o
c
a
p
ti
o
n
i
n
g
,
li
g
h
twe
ig
h
t
YO
LO,
m
u
lt
imo
d
a
l
c
o
n
tex
t
,
a
n
d
e
n
h
a
n
c
e
d
v
isi
o
n
c
o
rre
c
ti
o
n
.
H
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
c
d
1
1
1
5
@
s
e
o
g
a
n
g
.
a
c
.
k
r
.
S
iwo
o
B
y
u
n
re
c
e
iv
e
d
h
is
M
.
S
.
a
n
d
P
h
.
D.
d
e
g
re
e
s
in
c
o
m
p
u
ter
sc
ien
c
e
fro
m
KA
IS
T,
S
o
u
t
h
Ko
re
a
,
in
1
9
9
1
,
a
n
d
1
9
9
9
,
re
sp
e
c
ti
v
e
ly
.
S
in
c
e
M
a
rc
h
2
0
0
0
,
h
e
h
a
s
b
e
e
n
a
p
ro
fe
ss
o
r
i
n
t
h
e
De
p
a
rtme
n
t
o
f
S
o
ftwa
re
,
An
y
a
n
g
Un
iv
e
rsity
,
S
o
u
th
Ko
re
a
.
He
h
o
ld
s
a
P
h
.
D.
d
e
g
re
e
with
a
sp
e
c
ializa
ti
o
n
i
n
d
a
tab
a
se
tran
sa
c
ti
o
n
a
n
a
l
y
sis.
His
re
se
a
rc
h
a
re
a
s
a
re
Io
T
d
a
ta
p
ro
c
e
ss
in
g
,
I
o
T
m
a
c
h
in
e
lea
rn
in
g
,
e
m
b
e
d
d
e
d
sy
ste
m
,
b
ig
d
a
tab
a
se
,
flas
h
m
e
m
o
ry
d
a
tab
a
se
,
m
e
d
ica
l
h
e
a
lt
h
c
a
re
a
n
d
se
n
so
r
p
a
tt
e
rn
re
c
o
g
n
it
io
n
.
He
is
a
re
c
ip
ien
t
o
f
b
e
st
re
se
a
rc
h
sc
h
o
lar
a
wa
rd
a
n
d
b
e
st
p
a
p
e
r
a
wa
rd
.
His
n
e
w
re
se
a
r
c
h
in
tere
sts
in
c
lu
d
e
e
n
h
a
n
c
e
d
ima
g
e
/sig
n
a
l
p
ro
c
e
ss
in
g
,
d
e
e
p
lea
rn
in
g
,
m
u
lt
i
m
o
d
a
l,
fe
d
e
ra
ted
lea
rn
i
n
g
,
o
n
d
e
v
ice
AI,
sm
a
rt
g
las
s,
a
n
d
ro
a
d
fa
c
il
it
y
re
c
o
g
n
it
i
o
n
,
I
o
T
st
o
ra
g
e
s
y
ste
m
,
a
n
d
c
o
m
p
o
u
n
d
se
n
s
o
r
o
ff
l
o
a
d
in
g
.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
sw
b
y
u
n
@
a
n
y
a
n
g
.
a
c
.
k
r
.
Evaluation Warning : The document was created with Spire.PDF for Python.