I
nte
rna
t
io
na
l J
o
urna
l o
f
I
nfo
rm
a
t
ics a
nd
Co
m
m
un
ica
t
io
n T
ec
hn
o
lo
g
y
(
I
J
-
I
CT
)
Vo
l.
14
,
No
.
3
,
Dec
em
b
er
20
25
,
p
p
.
1
1
0
8
~
1
118
I
SS
N:
2252
-
8
7
7
6
,
DOI
:
1
0
.
1
1
5
9
1
/iji
ct
.
v14
i
3
.
pp
1
1
0
8
-
1
1
1
8
1108
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ict.
ia
esco
r
e.
co
m
Rev
o
lutionizin
g
h
uma
n activ
ity rec
o
g
nition with
pro
phet
a
lg
o
rithm and de
ep learning
J
a
y
k
um
a
r
S.
Dha
g
e
1
,
Av
ina
s
h K
.
G
ulv
e
2
1
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
D
r
.
B
a
b
a
s
a
h
e
b
A
mb
e
d
k
a
r
M
a
r
a
t
h
w
a
d
a
U
n
i
v
e
r
s
i
t
y
,
A
u
r
a
n
g
a
b
a
d
,
I
n
d
i
a
2
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
A
p
p
l
i
c
a
t
i
o
n
,
G
o
v
e
r
n
me
n
t
C
o
l
l
e
g
e
o
f
En
g
i
n
e
e
r
i
n
g
,
A
u
r
a
n
g
a
b
a
d
,
I
n
d
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
No
v
1
1
,
2
0
2
4
R
ev
is
ed
Feb
4
,
2
0
2
5
Acc
ep
ted
J
u
n
9
,
2
0
2
5
Va
rio
u
s
in
d
u
str
ies
,
su
c
h
a
s
h
e
a
lt
h
c
a
re
a
n
d
su
r
v
e
il
lan
c
e
,
d
e
p
e
n
d
h
e
a
v
il
y
o
n
th
e
a
b
il
it
y
to
re
c
o
g
n
ize
h
u
m
a
n
a
c
ti
v
it
y
.
Th
e
“
h
u
m
a
n
a
c
ti
v
it
y
re
c
o
g
n
i
ti
o
n
(HA
R)
u
sin
g
sm
a
rtp
h
o
n
e
s
d
a
ta
se
t
”
c
a
n
b
e
fo
u
n
d
in
t
h
e
UCI
o
n
li
n
e
re
p
o
sito
r
y
a
n
d
in
c
l
u
d
e
s
a
c
c
e
lero
m
e
ter
a
n
d
g
y
ro
sc
o
p
e
re
a
d
in
g
s
re
c
o
rd
e
d
d
u
ri
n
g
a
v
a
riet
y
o
f
h
u
m
a
n
a
c
ti
v
it
ies
.
Th
e
a
c
c
e
lero
m
e
ter
a
n
d
g
y
r
o
sc
o
p
e
sig
n
a
ls
a
re
a
lso
su
b
jec
ted
to
a
b
a
n
d
-
p
a
ss
fil
ter
to
e
li
m
in
a
te
u
n
wa
n
ted
fre
q
u
e
n
c
ies
a
n
d
b
a
c
k
g
ro
u
n
d
n
o
i
se
.
Th
is
m
e
th
o
d
e
ffe
c
ti
v
e
ly
d
e
c
re
a
se
s
th
e
d
ime
n
sio
n
a
l
it
y
o
f
th
e
fe
a
tu
re
sp
a
c
e
wh
il
e
imp
ro
v
i
n
g
th
e
m
o
d
e
l'
s
a
c
c
u
ra
c
y
a
n
d
e
fficie
n
c
y
.
“
Co
n
v
o
l
u
ti
o
n
a
l
n
e
u
ra
l
n
e
two
r
k
s
(CNN
s)
”
a
n
d
“
l
o
n
g
sh
o
r
t
-
term
m
e
m
o
ry
(LS
T
M
)
”
n
e
two
rk
s
a
re
c
o
m
b
in
e
d
t
o
c
re
a
te
p
y
ra
m
i
d
a
l
d
il
a
ted
c
o
n
v
o
lu
ti
o
n
a
l
m
e
m
o
ry
n
e
two
r
k
(P
DCMN),
w
h
ich
is
th
e
fin
a
l
p
ro
p
o
sa
l.
Re
su
lt
s
fro
m
e
x
p
e
rime
n
ts d
e
m
o
n
stra
te
th
e
e
ffe
c
ti
v
e
n
e
ss
a
n
d
re
li
a
b
il
it
y
o
f
t
h
e
su
g
g
e
ste
d
m
e
th
o
d
,
d
e
m
o
n
stra
ti
n
g
it
s
p
o
ten
t
ial
fo
r
p
re
c
ise
a
n
d
e
ffe
c
ti
v
e
HA
R
in
a
c
tu
a
li
t
y
sc
h
e
m
e
s.
K
ey
w
o
r
d
s
:
B
an
d
p
ass
f
ilter
Featu
r
e
d
im
en
s
io
n
ality
R
ed
u
ctio
n
Hu
m
an
ac
tiv
ity
r
ec
o
g
n
itio
n
L
o
n
g
s
h
o
r
t
-
ter
m
m
e
m
o
r
y
Pro
p
h
et
alg
o
r
ith
m
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
:
Av
in
ash
K.
Gu
lv
e
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
an
d
Ap
p
licatio
n
,
Go
v
er
n
m
en
t Co
lleg
e
o
f
E
n
g
in
ee
r
i
n
g
Au
r
an
g
ab
a
d
,
I
n
d
ia
E
m
ail:
ak
g
u
lv
e@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
T
o
d
ay
,
h
u
m
a
n
ac
tiv
ity
d
etec
tio
n
an
d
r
ec
o
g
n
itio
n
is
a
u
n
iq
u
e
d
o
m
ain
o
f
r
esear
ch
.
Dete
ctin
g
in
d
iv
id
u
al
ac
tiv
ities
n
o
w
a
d
a
y
is
v
er
y
ea
s
y
[1
]
,
[
2
]
.
Alm
o
s
t
ev
er
y
p
er
s
o
n
h
as
o
wn
ed
a
s
m
ar
t
p
h
o
n
e
to
d
ay
[
3
]
.
A
v
ar
iety
o
f
s
en
s
o
r
s
av
ailab
le
in
th
e
s
m
ar
t p
h
o
n
e
m
ak
es it p
o
s
s
ib
le
to
d
etec
t th
e
h
u
m
an
ac
tiv
ity
in
an
ac
cu
r
ate
way
[
4
]
,
[
5
]
.
T
h
e
co
m
p
o
n
e
n
ts
lik
e
ac
ce
ler
o
m
eter
,
g
y
r
o
s
co
p
e,
m
icr
o
p
h
o
n
e
an
d
ca
m
e
r
a
av
ailab
le
in
th
e
s
m
ar
t
p
h
o
n
e
a
r
e
u
s
ef
u
l
to
e
x
tr
ac
t
th
e
r
eq
u
ir
ed
o
u
tco
m
e
f
o
r
r
ec
o
g
n
izin
g
th
e
ac
tiv
ity
[
6
]
,
[
7
]
.
T
h
ese
d
ev
ices
ar
e
lo
w
co
s
t
an
d
tak
e
less
en
er
g
y
to
wo
r
k
.
H
u
m
an
ac
tiv
ity
r
ec
o
g
n
itio
n
(
HAR
)
h
as
a
v
ast
s
co
p
e
as
f
a
r
as
th
e
ap
p
licatio
n
s
ar
e
co
n
ce
r
n
ed
,
lik
e
h
ea
lth
ca
r
e,
s
o
cial
n
etwo
r
k
s
,
s
af
ety
,
d
etec
tio
n
o
f
s
u
s
p
icio
u
s
ac
tiv
ities
,
tr
an
s
p
o
r
tatio
n
a
n
d
s
u
r
v
eillan
c
e
s
y
s
tem
s
[
8
]
,
[
9
]
.
I
t
is
also
u
s
ed
in
r
ec
o
g
n
izin
g
th
e
d
if
f
e
r
en
ce
b
etwe
en
th
e
y
o
u
n
g
er
an
d
o
ld
er
p
eo
p
le
ac
tiv
ities
[
1
0
]
,
[
1
1
]
.
T
h
e
a
r
tific
ial
in
tellig
en
ce
(
AI
)
an
d
I
o
T
ar
e
h
elp
in
g
HAR
in
a
b
etter
way
[
1
2
]
.
B
ec
au
s
e
o
f
th
ese
n
ew
co
m
p
o
n
e
n
ts
,
th
e
HAR
is
n
o
t
o
n
ly
b
ec
o
m
in
g
m
o
r
e
co
m
p
lex
b
u
t
also
it
is
b
ec
o
m
in
g
th
e
m
o
s
t
in
ter
esti
n
g
an
d
o
p
en
to
p
ic
f
o
r
th
e
r
esear
ch
er
s
[
1
3
]
-
[
1
5
]
.
T
h
e
v
a
r
io
u
s
m
ac
h
in
e
lear
n
in
g
(
ML
)
ap
p
r
o
ac
h
es a
r
e
u
s
ed
to
e
x
tr
ac
t th
e
ex
ac
t in
f
o
r
m
atio
n
an
d
to
r
ea
ch
at
th
e
co
n
clu
s
io
n
[
1
6
]
-
[
1
8
]
.
I
n
o
r
d
e
r
to
ch
ec
k
th
e
r
esear
c
h
in
HAR,
it
is
n
ec
es
s
ar
y
to
s
tu
d
y
th
e
v
a
r
io
u
s
ap
p
r
o
ac
h
es
u
s
ed
in
th
e
p
ast
[
1
9
]
-
[
2
1
]
.
M
u
r
alid
h
a
r
an
et
a
l.
[
2
1
]
g
iv
es
d
etailed
in
f
o
r
m
atio
n
ab
o
u
t
th
e
one
-
d
im
en
s
io
n
al
(
1
D)
co
n
v
o
l
u
tio
n
al
n
e
u
r
al
n
etwo
r
k
(
C
NN
)
ap
p
r
o
ac
h
.
T
h
e
d
if
f
e
r
e
n
t
ML
ap
p
r
o
ac
h
es
wer
e
im
p
le
m
en
ted
as
a
p
ar
t
o
f
th
e
HAR in
th
is
r
esear
ch
wo
r
k
.
1
D
co
n
v
o
lu
tio
n
ap
p
r
o
ac
h
p
r
o
p
o
s
ed
in
th
e
p
ap
er
g
iv
es a
v
ali
d
atio
n
ac
cu
r
ac
y
o
f
9
6
.
1
3
%.
All
th
e
alg
o
r
ith
m
s
wer
e
im
p
lem
en
ted
o
n
t
h
e
s
tan
d
ar
d
UC
I
d
ataset.
T
h
e
o
n
e
-
d
im
en
s
io
n
al
(
1
D)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
R
ev
o
lu
tio
n
iz
in
g
h
u
ma
n
a
ctivity
r
ec
o
g
n
itio
n
w
ith
p
r
o
p
h
et
a
lg
o
r
ith
m
a
n
d
d
ee
p
… (
J
a
yk
u
ma
r
S
.
Dh
a
g
e
)
1109
c
o
n
v
o
l
u
tio
n
ap
p
r
o
ac
h
is
u
s
ed
with
th
e
d
ata
en
h
an
ce
m
en
t
s
tr
ateg
y
af
ter
u
n
d
er
g
o
in
g
tr
ain
in
g
,
as
s
u
g
g
ested
in
ar
ticle
[
2
2
]
.
W
ith
th
is
ap
p
r
o
ac
h
,
th
e
ac
c
u
r
a
cy
was
in
cr
e
ased
an
d
th
e
f
ac
to
r
s
lik
e
d
e
lay
,
co
m
p
u
tatio
n
al
co
m
p
lex
ity
wer
e
r
e
d
u
ce
d
.
Her
e,
th
e
n
u
m
b
er
s
o
f
f
alse
p
o
s
it
iv
e
wer
e
also
r
ed
u
ce
d
.
Kir
a
n
y
az
et
a
l.
[
2
3
]
h
as
p
r
o
v
id
e
d
g
o
o
d
in
f
o
r
m
atio
n
a
b
o
u
t
C
NN
an
d
its
o
v
er
all
wo
r
k
in
g
.
T
h
e
au
th
o
r
s
h
av
e
co
n
s
id
er
ed
en
g
i
n
ee
r
in
g
ap
p
licatio
n
s
with
r
ec
en
t
ad
v
an
ce
m
en
ts
.
1
-
D
co
n
v
o
lu
tio
n
is
p
er
f
ec
t
to
tr
ain
a
n
d
e
v
en
with
m
in
im
u
m
co
m
p
u
tatio
n
al
c
o
m
p
lex
ity
,
it
g
iv
es
co
n
s
id
er
ab
le
a
cc
u
r
ac
y
i
n
v
ar
io
u
s
ap
p
licatio
n
s
.
T
h
r
ee
ac
tiv
ity
lab
els
s
u
ch
as
L
AYI
NG,
SIT
T
I
NG
,
AND
STAN
DI
NG
wer
e
co
n
s
id
er
e
d
f
r
o
m
HAR
d
ataset
u
s
e
b
y
Min
ar
n
o
et
a
l.
[
2
4
]
.
T
h
ey
h
a
v
e
im
p
lem
en
ted
all
th
e
p
r
e
v
io
u
s
ly
m
e
n
tio
n
e
d
alg
o
r
ith
m
s
alo
n
g
with
a
n
ew
ap
p
r
o
ac
h
as
a
g
r
ad
ie
n
t
b
o
o
s
t.
T
h
e
l
o
g
is
tic
r
eg
r
ess
io
n
(
LR
)
m
o
d
el,
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
e
(
SVM
)
,
an
d
lin
ea
r
k
er
n
el
m
o
d
el
h
av
e
all
b
ee
n
co
m
p
ar
ed
b
y
th
e
au
th
o
r
s
.
T
h
e
m
ax
im
u
m
ac
cu
r
ac
y
r
e
p
o
r
ted
b
y
SVM
with
r
ad
ial
b
asis
f
u
n
ctio
n
(
R
B
F
)
k
er
n
el
m
o
d
el
is
9
8
.
9
6
%.
C
h
en
et
a
l.
[
1
]
u
s
es
a
g
r
a
p
h
ics
p
r
o
c
ess
in
g
u
n
it
(
GPU
)
f
r
am
ewo
r
k
.
I
t
em
p
lo
y
s
a
1
-
D
co
n
v
o
l
u
tio
n
to
co
n
v
er
t
s
in
ce
th
e
tim
e
d
o
m
ain
to
th
e
f
r
eq
u
e
n
cy
f
ield
.
T
h
e
d
ataset
is
Mu
s
ic
-
Net.
I
t
ca
n
p
er
f
o
r
m
s
p
ec
tr
o
g
r
am
ex
tr
ac
tio
n
.
3
4
s
p
ec
tr
o
g
r
a
m
ty
p
es
with
v
ar
io
u
s
p
ar
am
eter
s
m
ay
b
e
ex
tr
ac
ted
in
2
.
8
h
o
u
r
s
.
Z
h
an
g
et
a
l.
[
2
5
]
o
p
er
ated
a
t
y
p
ical
8
-
la
y
er
C
NN,
with
b
at
ch
n
o
r
m
aliza
tio
n
a
n
d
d
r
o
p
o
u
t.
B
y
m
er
g
i
n
g
“
two
-
lay
er
g
r
ap
h
co
n
v
o
lu
tio
n
al
n
et
wo
r
k
(
GC
N
)
”
,
t
h
e
“
B
DR
-
C
N
N
-
GC
N
”
is
cr
ea
ted
.
L
ee
et
a
l.
[
4
]
p
r
o
p
o
s
es
a
1
-
D
c
o
n
v
o
l
u
tio
n
f
o
r
th
e
class
if
icatio
n
o
f
a
ctiv
ities
r
ec
o
g
n
ized
b
y
th
e
m
o
d
el.
I
n
th
at
m
o
d
el
th
r
e
e
ac
tiv
ities
n
am
ely
r
u
n
,
walk
,
an
d
s
till
ar
e
im
p
lem
en
ted
.
T
h
e
ac
cu
r
ac
y
ac
h
iev
ed
with
th
is
alg
o
r
ith
m
is
9
2
.
7
1
%.
I
n
th
e
p
r
o
p
o
s
ed
wo
r
k
,
a
p
ar
t
f
r
o
m
v
ar
io
u
s
ML
alg
o
r
ith
m
s
s
u
ch
as
L
R
,
SVM,
r
an
d
o
m
f
o
r
est
(
R
F)
,
d
ec
is
io
n
tr
ee
(
DT
)
,
n
eu
r
a
l
n
etwo
r
k
(
NN)
,
th
e
d
ee
p
lear
n
i
n
g
ap
p
r
o
ac
h
es
ar
e
also
im
p
le
m
en
ted
.
A
n
ew
d
ataset
is
cr
ea
ted
with
th
e
h
elp
o
f
an
s
en
s
o
r
ap
p
an
d
th
en
all
th
e
d
ee
p
lear
n
in
g
ap
p
r
o
ac
h
es
ar
e
im
p
lem
en
ted
.
T
h
e
p
r
o
p
h
et
alg
o
r
ith
m
s
with
th
e
co
m
b
in
atio
n
o
f
T
DL
an
d
lo
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
(
L
STM
)
ar
e
g
iv
in
g
g
o
o
d
r
esu
lts
as
co
m
p
ar
ed
t
o
all
o
t
h
er
alg
o
r
ith
m
s
.
T
h
i
s
s
t
u
d
y
i
n
v
e
s
t
i
g
a
t
e
d
t
h
e
e
f
f
e
c
t
i
v
e
n
es
s
o
f
a
d
v
a
n
c
e
d
d
e
e
p
l
e
a
r
n
i
n
g
a
l
g
o
r
i
t
h
m
s
f
o
r
H
A
R
u
s
i
n
g
s
m
a
r
t
p
h
o
n
e
s
e
n
s
o
r
d
a
t
a
.
W
h
i
le
e
a
r
l
i
e
r
s
t
u
d
i
es
h
a
v
e
e
x
p
l
o
r
ed
t
h
e
i
m
p
a
c
t
o
f
t
r
a
d
i
t
i
o
n
a
l
ML
t
e
c
h
n
i
q
u
e
s
s
u
c
h
as
LR
,
SV
M
,
DT
,
a
n
d
RF
o
n
HA
R
,
t
h
e
y
h
a
v
e
n
o
t
e
x
p
l
i
ci
t
l
y
ad
d
r
e
s
s
e
d
t
h
e
p
o
t
e
n
ti
a
l
o
f
d
e
e
p
l
e
a
r
n
i
n
g
m
o
d
e
l
s
t
o
h
a
n
d
l
e
c
o
m
p
l
e
x
t
e
m
p
o
r
a
l
p
a
t
te
r
n
s
a
n
d
l
a
r
g
e
-
s
c
a
le
d
a
t
as
e
ts
ef
f
e
c
t
i
v
e
l
y
.
F
u
r
t
h
e
r
m
o
r
e
,
p
r
i
o
r
r
e
s
e
a
r
c
h
o
f
t
e
n
r
e
l
i
ed
o
n
l
i
m
i
t
e
d
d
at
as
e
ts
wi
t
h
c
o
n
s
tr
a
i
n
e
d
a
c
t
i
v
i
t
y
l
a
b
el
s
,
l
a
c
k
i
n
g
d
i
v
e
r
s
i
t
y
a
n
d
r
e
a
l
-
w
o
r
l
d
a
p
p
l
ic
a
b
i
l
i
t
y
.
T
h
i
s
s
t
u
d
y
a
i
m
s
t
o
b
r
i
d
g
e
t
h
es
e
g
a
p
s
b
y
i
n
t
r
o
d
u
c
i
n
g
r
o
b
u
s
t
d
e
e
p
l
e
a
r
n
i
n
g
m
o
d
e
l
s
,
T
D
L
-
L
S
T
M
a
n
d
L
S
T
M
-
T
D
L
,
t
r
a
i
n
e
d
o
n
a
c
o
m
p
r
e
h
e
n
s
i
v
e
d
a
t
as
e
t
w
i
t
h
s
i
x
a
c
t
i
v
it
y
l
a
b
e
ls
,
a
n
d
e
v
a
l
u
a
ti
n
g
t
h
e
i
r
p
e
r
f
o
r
m
a
n
c
e
u
n
d
e
r
r
e
a
l
-
w
o
r
l
d
c
o
n
d
i
t
i
o
n
s
.
W
e
f
o
u
n
d
th
at
t
h
e
p
r
o
p
o
s
ed
d
ee
p
lear
n
in
g
m
o
d
els,
T
DL
-
L
STM
an
d
L
STM
-
T
DL
,
s
ig
n
if
ican
tly
o
u
tp
er
f
o
r
m
ed
tr
ad
itio
n
al
ML
m
eth
o
d
s
in
HAR
,
ac
h
iev
in
g
ac
cu
r
ac
ies
o
f
9
8
.
4
9
%
a
n
d
9
8
.
0
9
%,
r
esp
ec
tiv
ely
.
T
h
e
r
esu
lts
in
d
icate
d
th
at
th
ese
m
o
d
els
ef
f
ec
tiv
el
y
ca
p
tu
r
ed
co
m
p
lex
tem
p
o
r
al
p
atter
n
s
a
n
d
s
u
b
tle
v
ar
iatio
n
s
in
th
e
d
ata.
T
h
e
T
DL
-
L
STM
m
o
d
el
d
em
o
n
s
tr
ated
a
n
o
tab
ly
h
ig
h
e
r
p
r
ec
is
io
n
a
n
d
r
ec
all
f
o
r
ac
tiv
ities
lik
e
“
walk
in
g
”
an
d
“stan
d
in
g
”
c
o
m
p
ar
ed
to
o
th
er
m
eth
o
d
s
,
s
u
g
g
esti
n
g
its
r
o
b
u
s
tn
ess
in
r
ec
o
g
n
izin
g
d
y
n
am
ic
an
d
s
tatic
ac
tiv
ities
.
T
h
ese
f
in
d
in
g
s
u
n
d
e
r
s
co
r
e
th
e
ab
ilit
y
o
f
d
ee
p
lear
n
in
g
a
r
ch
itectu
r
es
to
p
r
o
ce
s
s
lar
g
e
-
s
ca
le
d
atasets
an
d
d
eliv
er
s
u
p
er
i
o
r
p
er
f
o
r
m
a
n
ce
in
HAR task
s
.
2.
G
E
NE
R
AL
AR
CH
I
T
E
C
T
U
RE
/PR
O
C
E
DUR
E
Fig
u
r
e
1
s
h
o
ws
th
e
p
r
o
ce
s
s
f
lo
w
f
o
r
th
e
HAR
is
g
iv
en
.
T
o
co
llect
th
e
d
ata
f
o
r
th
e
ab
o
v
e
f
lo
w,
an
ap
p
is
d
ev
elo
p
ed
.
T
h
e
d
ata
is
co
ll
ec
ted
f
r
o
m
it.
A
s
m
ar
tp
h
o
n
e
R
ed
m
i
No
te
9
p
r
o
is
u
s
ed
f
o
r
th
e
d
ata
c
o
llectio
n
.
T
h
er
e
ar
e
s
ix
lab
els
av
ailab
le,
an
d
ea
ch
p
er
s
o
n
h
as
a
s
m
ar
tp
h
o
n
e.
T
h
r
ee
-
a
x
ial
lin
ea
r
ac
ce
ler
atio
n
an
d
th
r
ee
-
ax
ial
an
g
u
lar
v
elo
city
wer
e
r
e
co
r
d
ed
with
t
h
e
ass
is
tan
ce
o
f
th
e
b
u
ilt
-
in
ac
ce
ler
o
m
eter
a
n
d
g
y
r
o
s
co
p
e.
T
h
e
r
e
ar
e
s
ix
ac
tiv
ity
la
b
els.
All
th
e
r
esu
lts
ar
e
v
i
d
eo
-
r
ec
o
r
d
e
d
t
o
g
et
t
h
e
lab
els.
T
h
e
d
ata
s
et
f
o
r
m
ed
th
r
o
u
g
h
th
e
ex
p
er
im
en
t sam
p
les ar
e
s
p
lit in
to
two
s
ets,
wh
er
e
o
n
e
s
et
with
7
0
% o
f
s
am
p
les is
u
s
ed
as th
e
tr
ain
in
g
d
ata
an
d
3
0
%
as
th
e
test
d
ata.
T
h
e
p
r
ep
r
o
ce
s
s
in
g
o
f
s
en
s
o
r
s
ig
n
als
is
d
o
n
e
b
y
ap
p
ly
i
n
g
f
ilter
s
.
T
h
en
,
th
ey
ar
e
s
am
p
le
d
with
s
lid
in
g
win
d
o
ws
th
at
o
v
er
lap
b
y
5
0
%
an
d
last
2
.
5
6
s
ec
o
n
d
s
.
Featu
r
e
v
e
cto
r
s
ar
e
g
en
er
ated
f
r
o
m
ea
ch
tim
e
win
d
o
w
b
y
ca
lcu
latin
g
m
etr
ics
ac
r
o
s
s
two
d
o
m
ain
s
:
tr
iax
ial
ac
ce
ler
atio
n
,
wh
ich
ac
co
u
n
ts
f
o
r
b
o
d
y
ac
ce
ler
atio
n
,
an
d
tr
iax
ial
an
g
u
lar
v
elo
city
.
T
h
is
p
r
o
ce
s
s
y
iel
d
s
a
to
tal
o
f
5
6
1
attr
ib
u
tes
f
o
r
an
aly
s
is
.
T
h
e
d
ata
is
ca
teg
o
r
ized
in
to
s
ix
s
p
ec
if
ic
ac
tiv
ity
lab
els:
0
r
ep
r
esen
ts
lay
in
g
,
1
co
r
r
esp
o
n
d
s
to
s
itti
n
g
,
2
in
d
icate
s
s
tan
d
in
g
,
3
is
f
o
r
walk
in
g
,
4
d
en
o
tes
walk
in
g
d
o
wn
s
tair
s
,
an
d
5
is
f
o
r
walk
in
g
u
p
s
tair
s
.
Fu
r
th
e
r
m
o
r
e,
ea
c
h
f
ea
tu
r
e
v
ec
to
r
is
tag
g
ed
with
a
n
id
en
tifie
r
to
s
p
ec
if
y
th
e
s
u
b
ject.
Pre
-
p
r
o
ce
s
s
in
g
:
i
t
p
u
r
p
o
s
es
to
r
aise
th
e
d
ata'
s
q
u
ality
s
o
th
at
ev
e
r
y
o
n
e
ca
n
an
aly
z
e
it
m
o
r
e
ef
f
ec
tiv
ely
.
I
n
th
is
r
esear
ch
wo
r
k
,
s
lid
in
g
win
d
o
w
s
eg
m
e
n
tatio
n
is
u
s
ed
to
b
r
ea
k
u
p
th
e
co
n
tin
u
o
u
s
s
en
s
o
r
d
ata
in
to
s
h
o
r
ter
tim
e
win
d
o
w
s
o
r
ch
u
n
k
s
o
f
a
s
et
len
g
th
.
B
an
d
-
p
ass
f
ilter
is
u
s
ed
to
f
ilter
th
e
ac
ce
ler
o
m
eter
an
d
g
y
r
o
s
co
p
e
s
ig
n
als to
elim
i
n
ate
u
n
wan
ted
f
r
eq
u
en
cies a
n
d
n
o
is
e.
Min
-
m
a
x
s
ca
lin
g
is
u
s
ed
to
n
o
r
m
alize
th
e
s
ig
n
als an
d
p
l
ac
e
th
em
with
in
a
p
r
ed
eter
m
in
e
d
r
an
g
e
f
o
r
co
n
s
is
ten
t c
o
m
p
ar
is
o
n
an
d
an
aly
s
is
.
B
an
d
p
ass
f
ilter
:
it
k
ee
p
s
th
e
r
e
d
u
ce
e
x
ten
t
b
an
d
o
f
f
r
eq
u
e
n
cies
b
y
elim
in
ate
th
e
v
er
y
lo
w
f
r
e
q
u
en
c
y
a
n
d
v
er
y
h
ig
h
f
r
eq
u
en
cy
p
ar
t.
E
d
g
es a
r
e
im
p
r
o
v
ed
w
h
ile
n
o
is
e
is
also
d
im
in
is
h
ed
u
s
in
g
b
an
d
p
ass
f
ilter
in
g
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
7
7
6
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
,
Vo
l.
14
,
No
.
3
,
Dec
em
b
er
20
25
:
1
1
0
8
-
1
1
1
8
1110
Fig
u
r
e
1
.
Ar
c
h
itectu
r
al
d
iag
r
a
m
Min
m
ax
s
ca
lin
g
:
th
is
s
im
p
l
e
n
o
r
m
aliza
tio
n
m
eth
o
d
y
iel
d
s
th
e
co
m
m
o
n
n
u
m
er
ical
r
a
n
g
e
o
f
th
e
s
co
r
es
[
0
,
1
]
wh
ile
also
m
ai
n
tain
in
g
th
e
o
r
ig
in
al
d
is
tr
ib
u
tio
n
s
h
ap
e
s
with
th
e
ex
ce
p
tio
n
o
f
a
s
ca
lin
g
f
ac
t
o
r
.
L
et
s
tan
d
f
o
r
a
g
r
o
u
p
o
f
r
aw
m
atc
h
r
esu
lts
o
u
t
o
f
a
ce
r
tain
m
atc
h
er
.
T
h
e
n
,
′
r
ep
r
esen
ts
th
e
ad
ju
s
ted
s
co
r
e
o
f
.
Ass
u
m
in
g
th
at
(
)
an
d
(
)
r
ep
r
esen
t
th
e
u
tm
o
s
t
an
d
m
in
im
u
m
s
tan
d
ar
d
s
f
o
r
th
e
r
aw
m
atc
h
in
g
s
co
r
es.
T
h
e
“
n
o
r
m
alize
d
s
co
r
e
”
is
th
en
co
n
s
id
er
ed
in
(
1
)
.
′
=
(
−
(
)
)
/
(
)
−
(
)
(
1
)
Du
e
to
its
h
ig
h
s
u
s
ce
p
tib
ilit
y
t
o
o
u
tlier
s
in
th
e
esti
m
atio
n
d
ata,
th
is
m
eth
o
d
is
n
o
t
r
o
b
u
s
t.
Du
e
to
th
e
ex
is
ten
ce
o
f
o
u
tlier
s
,
th
e
m
ajo
r
ity
o
f
th
e
d
ata
is
co
n
ce
n
tr
ated
o
n
ly
with
i
n
a
s
m
aller
r
an
g
e
.
Featu
r
e
ex
tr
ac
tio
n
:
it
h
elp
s
to
tak
e
o
u
t
th
e
b
est
f
ea
tu
r
e
f
r
o
m
th
o
s
e
en
o
r
m
o
u
s
d
ata
s
ets
b
y
p
ick
an
d
co
m
b
in
in
g
v
ar
iab
les
in
to
tr
aits
.
I
n
th
is
r
ese
ar
ch
wo
r
k
,
s
tatis
tical
f
ea
tu
r
es,
co
r
r
elatio
n
an
aly
s
is
,
an
d
PS
D
ar
e
u
s
ed
to
ex
tr
ac
t th
e
f
ea
tu
r
es.
Statis
t
ical
f
ea
tu
r
e:
th
e
tech
n
iq
u
e
o
f
clu
s
ter
in
g
an
d
ev
alu
atin
g
d
ata
f
o
r
th
e
g
o
al
o
f
d
eter
m
in
e
o
r
ig
in
als
an
d
tr
en
d
s
is
k
n
o
wn
as
s
tatis
ti
ca
l
an
aly
s
is
.
I
t
is
an
ap
p
r
o
ac
h
f
o
r
er
ad
icate
b
ias
f
r
o
m
d
ata
v
alu
atio
n
b
y
m
ea
n
s
o
f
n
u
m
er
ical
r
ev
iew.
T
h
is
ap
p
r
o
ac
h
is
p
r
o
f
itab
le
f
o
r
ass
em
b
lag
e
r
esear
ch
i
n
ter
p
r
etatio
n
s
,
s
et
u
p
s
tatis
tical
m
o
d
els,
an
d
p
lan
n
in
g
i
n
v
esti
g
atio
n
s
an
d
r
esear
ch
.
I
n
(
2
)
an
d
(
3
)
d
esig
n
ate
th
e
m
at
h
em
atica
l
m
o
d
el
o
f
m
ea
n
an
d
s
tan
d
ar
d
d
ev
iatio
n
.
Mean
,
=
∑
−
1
=
0
(
)
(
2
)
Stan
d
ar
d
d
e
v
iatio
n
,
=
√
1
∑
=
1
(
−
)
2
(
3
)
Min
im
u
m
:
t
h
e
m
in
im
u
m
n
u
m
b
er
in
o
u
r
s
et
o
f
d
ata
is
th
e
d
at
a
v
alu
e
th
at
is
le
s
s
th
an
o
r
ca
p
ab
le
o
f
all
o
th
er
v
alu
es.
I
f
all
o
f
o
u
r
d
ata
wer
e
ar
r
an
g
ed
in
i
n
cr
ea
s
in
g
o
r
d
er
o
f
im
p
o
r
tan
ce
,
th
e
lo
west
n
u
m
b
er
o
n
o
u
r
lis
t
wo
u
ld
b
e
it.
T
h
e
m
in
im
u
m
v
alu
e
m
ay
o
cc
u
r
m
o
r
e
th
an
o
n
ce
in
th
e
d
ata
s
et,
b
u
t it
is
s
til
l a
e
x
ce
p
tio
n
al
n
u
m
b
er
b
y
d
ef
i
n
itio
n
.
T
h
er
e
ab
le
b
e
t
wo
m
in
im
a
s
in
ce
o
n
l
y
o
n
e
o
f
t
h
es
e
v
alu
es c
an
b
e
g
r
ea
ter
th
a
n
th
e
o
th
e
r
.
Ma
x
im
u
m
:
t
h
e
v
alu
e
th
at
e
x
ce
ed
s
o
r
is
o
n
a
p
ar
with
e
v
er
y
o
th
e
r
v
alu
e
in
th
e
s
et
o
f
d
ata
is
co
n
s
id
er
ed
to
b
e
its
m
ax
im
u
m
v
alu
e.
I
f
all
o
f
o
u
r
d
ata
wer
e
ar
r
an
g
e
d
in
ascen
d
in
g
o
r
d
er
,
t
h
e
h
ig
h
est
n
u
m
b
e
r
wo
u
ld
b
e
th
e
las
t o
n
e
lis
ted
.
T
h
e
g
r
ea
test
is
a
s
in
g
le
n
u
m
b
e
r
f
o
r
a
p
a
r
ticu
lar
s
et
o
f
f
ac
ts
.
Fo
r
a
d
ata
s
et,
th
er
e
is
o
n
ly
o
n
e
m
ax
im
u
m
,
b
u
t
th
is
n
u
m
b
er
m
a
y
b
e
r
ep
ea
ted
.
T
h
er
e
ca
n
n
ev
er
b
e
two
m
ax
im
a
s
in
ce
th
er
e
wo
u
ld
alwa
y
s
b
e
o
n
e
v
al
u
e
th
at
is
g
r
e
ater
th
an
th
e
o
th
er
.
Sk
e
wn
ess
:
d
escr
ib
es
th
e
u
n
ev
en
n
ess
o
r
d
is
to
r
tio
n
p
r
esen
t
in
a
s
tatis
tical
d
is
tr
ib
u
tio
n
.
T
h
is
o
cc
u
r
s
wh
en
d
ata
p
o
in
ts
d
o
n
'
t
s
p
r
ea
d
o
u
t
s
y
m
m
etr
ically
o
n
eith
er
s
id
e
o
f
th
e
m
ed
ian
in
wh
at
wo
u
ld
ty
p
ically
b
e
a
b
ell
-
s
h
ap
ed
cu
r
v
e.
W
h
en
th
e
b
ell
cu
r
v
e
lean
s
to
o
n
e
s
id
e
-
ei
th
er
lef
t
o
r
r
ig
h
t
-
it
s
ig
n
als
th
at
th
e
d
is
tr
ib
u
tio
n
is
u
n
b
alan
ce
d
.
I
n
(
4
)
r
ep
r
esen
ts
t
h
e
m
ath
em
atica
l e
q
u
atio
n
o
f
s
k
ewn
ess
.
3
=
−
3
∑
−
1
=
0
(
−
)
3
(
)
(
4
)
Ku
r
to
s
is
:
is
a
m
ea
s
u
r
em
en
t
o
f
th
e
o
u
tlier
n
atu
r
e
o
f
a
r
ea
l
-
v
alu
ed
r
an
d
o
m
v
a
r
iab
le'
s
p
r
o
b
ab
i
lity
d
is
tr
ib
u
tio
n
.
I
n
(
5
)
r
e
p
r
esen
ts
th
e
m
ath
em
atic
al
eq
u
atio
n
o
f
k
u
r
to
s
is
.
4
=
−
4
∑
−
1
=
0
(
−
)
4
(
)
−
3
(
5
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
R
ev
o
lu
tio
n
iz
in
g
h
u
ma
n
a
ctivity
r
ec
o
g
n
itio
n
w
ith
p
r
o
p
h
et
a
lg
o
r
ith
m
a
n
d
d
ee
p
… (
J
a
yk
u
ma
r
S
.
Dh
a
g
e
)
1111
E
n
er
g
y
:
is
also
r
ef
er
r
ed
to
as
th
e
u
n
if
o
r
m
ity
o
r
an
g
u
la
r
s
ec
o
n
d
m
o
m
e
n
t.
I
n
(
6
)
r
ep
r
esen
t
s
th
e
m
ath
em
atica
l
eq
u
atio
n
o
f
en
e
r
g
y
.
=
∑
−
1
=
0
[
(
)
]
2
(
6
)
T
h
e
m
eth
o
d
s
ec
tio
n
g
iv
en
b
e
lo
w
o
u
tlin
es
t
h
e
s
y
s
tem
atic
a
p
p
r
o
ac
h
u
s
ed
to
d
e
v
elo
p
an
d
ev
alu
ate
ad
v
an
ce
d
d
ee
p
lear
n
in
g
m
o
d
e
ls
f
o
r
HAR
.
Data
wer
e
c
o
llected
u
s
in
g
a
s
m
ar
tp
h
o
n
e
a
p
p
lic
atio
n
th
at
r
ec
o
r
d
ed
ac
ce
ler
o
m
eter
an
d
g
y
r
o
s
co
p
e
r
ea
d
in
g
s
d
u
r
in
g
s
ix
p
r
ed
ef
in
ed
ac
tiv
ities
:
lay
in
g
,
s
itti
n
g
,
s
tan
d
in
g
,
walk
in
g
,
walk
in
g
d
o
wn
s
tair
s
,
an
d
walk
in
g
u
p
s
tair
s
.
Pre
p
r
o
ce
s
s
in
g
in
clu
d
ed
s
lid
in
g
win
d
o
w
s
eg
m
e
n
tatio
n
,
b
an
d
-
p
ass
f
ilter
in
g
to
r
e
m
o
v
e
n
o
is
e,
an
d
m
in
-
m
a
x
s
ca
lin
g
f
o
r
n
o
r
m
aliza
tio
n
,
r
esu
ltin
g
in
5
6
1
f
e
atu
r
e
attr
ib
u
tes
p
e
r
win
d
o
w.
Statis
tical
f
ea
tu
r
e
ex
tr
ac
tio
n
tech
n
iq
u
es,
s
u
ch
as
m
ea
n
,
s
tan
d
ar
d
d
ev
iatio
n
,
s
k
e
wn
ess
,
an
d
k
u
r
to
s
is
,
wer
e
ap
p
lied
to
en
h
a
n
ce
d
ata
r
ep
r
esen
tatio
n
.
T
wo
d
ee
p
lear
n
in
g
m
o
d
els,
T
DL
-
L
STM
an
d
L
STM
-
T
DL
,
wer
e
tr
ain
ed
o
n
a
7
0
-
3
0
tr
ain
-
test
s
p
lit
u
s
in
g
th
e
Ker
as
f
r
am
e
wo
r
k
.
T
h
eir
p
er
f
o
r
m
an
ce
was
co
m
p
ar
ed
ag
ai
n
s
t
tr
ad
itio
n
al
ML
m
o
d
els
u
s
in
g
m
etr
ics
lik
e
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
-
s
co
r
e
to
v
alid
ate
th
eir
ef
f
ec
tiv
en
ess
.
3.
P
RO
P
O
SE
D
M
E
T
H
O
D
T
h
e
p
r
o
p
o
s
ed
m
eth
o
d
o
l
o
g
y
o
f
p
r
o
p
h
et
alg
o
r
ith
m
is
s
h
o
wn
i
n
Fig
u
r
e
2
.
Fo
r
th
e
d
ata
in
th
e
tim
e
r
an
g
e
lik
e
s
to
ck
p
r
ices
an
d
v
id
eo
f
r
a
m
es,
L
STM
is
th
e
p
er
f
ec
t
ap
p
r
o
ac
h
.
L
STM
s
elec
tiv
ely
r
em
e
m
b
er
s
th
e
p
atter
n
s
f
o
r
l
o
n
g
d
u
r
atio
n
s
o
f
tim
e.
I
t
h
as
b
o
th
th
e
o
p
tio
n
s
lik
e,
k
ee
p
in
g
th
e
t
h
in
g
s
in
m
em
o
r
y
f
o
r
s
h
o
r
t
o
r
l
o
n
g
d
u
r
atio
n
o
f
tim
e.
I
t
is
an
im
p
r
o
v
em
en
t
o
v
e
r
r
ec
u
r
r
e
n
t
n
eu
r
al
n
etwo
r
k
(
R
NN)
.
I
n
L
STM
,
t
h
er
e
ar
e
ce
ll
s
tate
s
with
d
if
f
er
en
t
d
ep
en
d
en
cies.
Her
e,
th
ey
d
o
n
o
t
m
an
i
p
u
lat
e
th
e
en
tire
in
f
o
r
m
atio
n
b
u
t
th
ey
m
o
d
if
y
th
e
m
s
lig
h
tly
.
L
STM
s
ar
e
ess
en
tial
in
s
eq
u
en
ce
class
if
icatio
n
b
ec
au
s
e
th
ey
ca
n
ef
f
ec
tiv
ely
lear
n
p
atter
n
s
d
ir
ec
tl
y
f
r
o
m
r
aw
ti
m
e
s
er
ies
d
ata.
T
h
is
elim
in
ates
th
e
n
ee
d
f
o
r
m
an
u
al
f
ea
tu
r
e
en
g
in
ee
r
in
g
,
all
o
win
g
th
e
m
o
d
el
to
p
er
f
o
r
m
well
with
o
u
t
r
ely
in
g
o
n
d
o
m
ain
-
s
p
ec
if
ic
ex
p
er
tis
e.
I
t
ac
h
iev
es
eq
u
iv
alen
t
s
p
e
ed
an
d
ca
n
q
u
ick
ly
in
ter
n
alize
a
tim
e
s
er
ies
d
ata
f
o
r
m
at.
T
o
ad
o
p
t
ea
c
h
in
p
u
t
b
ef
o
r
e
o
r
af
ter
th
is
L
STM
lay
er
,
tim
e
d
is
tr
ib
u
ted
lay
er
ap
p
r
o
ac
h
ca
n
b
e
v
er
y
u
s
ef
u
l.
Ker
as
p
r
o
v
id
es
a
n
ice
s
o
lu
tio
n
f
o
r
th
e
d
ata
wh
er
e
f
r
am
e
b
y
f
r
am
e
ac
tiv
ities
,
co
n
s
ec
u
tiv
e
an
d
s
eq
u
en
tial
ac
tio
n
s
ar
e
av
ailab
le.
I
t
is
n
am
ed
as
“
tim
e
d
i
s
tr
ib
u
ted
”
lay
er
.
S
o
,
with
th
is
lay
er
an
d
with
th
e
L
STM
,
th
is
m
o
d
el
ca
n
p
r
o
v
id
e
v
er
y
ac
cu
r
ate
r
esu
lts
,
p
r
o
v
id
ed
th
at
th
e
f
i
n
e
s
y
n
ch
r
o
n
izatio
n
is
ac
h
iev
ed
.
Fig
u
r
e
2
.
Pro
p
h
et
alg
o
r
ith
m
(
L
STM
-
T
DM
/TD
M
-
L
STM
)
T
im
e
d
is
tr
ib
u
ted
la
y
er
wo
r
k
with
s
ev
er
al
in
p
u
ts
.
I
t
p
r
o
d
u
ce
s
o
n
e
o
u
tp
u
t
p
er
i
n
p
u
t
t
o
g
et
t
h
e
r
esu
lt
i
n
tim
e.
I
n
HAR,
it
is
r
eq
u
ir
ed
to
ch
ec
k
an
o
b
ject
in
m
o
tio
n
.
So
,
b
ef
o
r
e
d
etec
tin
g
th
e
m
o
v
e
m
en
t,
s
ea
r
ch
in
g
th
e
o
b
ject
is
m
o
r
e
im
p
o
r
ta
n
t.
T
h
at
is
wh
y
,
in
th
is
m
o
d
el,
it
n
ee
d
s
to
m
ak
e
co
n
v
o
lu
tio
n
s
b
ef
o
r
e
L
STM
.
T
im
e
d
is
tr
ib
u
ted
lay
er
is
v
er
y
s
tr
o
n
g
in
th
e
s
en
s
e
th
at,
ir
r
esp
ec
tiv
e
o
f
its
p
o
s
itio
n
with
L
STM
,
t
h
e
ef
f
ec
t
o
n
th
e
d
ata
will b
e
s
am
e.
On
ly
o
n
e
m
o
d
el
ca
n
d
o
t
h
e
wo
r
k
.
Ker
as is
u
s
ed
to
m
ak
e
m
o
v
em
e
n
t p
r
ed
ictio
n
an
d
r
ec
o
g
n
itio
n
.
L
STM
:
R
NN
ar
ch
itectu
r
es
wi
th
L
S
T
M
ca
n
h
an
d
le
s
eq
u
en
ti
al
d
ata
with
lo
n
g
-
d
is
tan
ce
d
ep
en
d
en
cies.
L
STM
s
in
co
r
p
o
r
ate
m
em
o
r
y
c
ells
an
d
s
p
ec
ialized
g
atin
g
m
e
ch
an
is
m
s
,
in
co
n
tr
ast
to
co
n
v
e
n
tio
n
al
R
NNs.
As
a
r
esu
lt,
th
e
n
etwo
r
k
ca
n
elec
tiv
ely
r
em
em
b
er
o
r
f
o
r
g
et
i
n
f
o
r
m
atio
n
ev
en
tu
ally
,
ef
f
ec
tiv
el
y
ca
p
tu
r
in
g
an
d
p
r
eser
v
in
g
s
ig
n
if
ican
t
tem
p
o
r
al
p
atter
n
s
.
T
h
ese
g
ates
co
n
tr
o
l
th
e
f
lo
w
o
f
in
f
o
r
m
atio
n
with
in
th
e
n
etwo
r
k
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
7
7
6
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
,
Vo
l.
14
,
No
.
3
,
Dec
em
b
er
20
25
:
1
1
0
8
-
1
1
1
8
1112
L
STM
h
as
d
em
o
n
s
tr
ated
s
u
cc
ess
in
a
ca
teg
o
r
y
o
f
task
s
,
as
well
as
s
p
ee
ch
r
ec
o
g
n
itio
n
,
la
n
g
u
ag
e
tr
an
s
latio
n
,
an
d
s
en
tim
en
t a
n
aly
s
is
,
wh
e
r
e
it is
ess
en
tia
l to
co
m
p
r
eh
e
n
d
a
n
d
m
o
d
el
s
eq
u
en
tial d
ata.
Fu
lly
co
n
n
ec
ted
lay
er
:
t
h
e
d
en
s
e
lay
er
,
o
f
ten
r
ef
er
r
e
d
to
as
t
h
e
f
u
lly
co
n
n
ec
te
d
lay
er
in
NN
s
,
p
lay
s
a
k
ey
r
o
le
in
th
e
a
r
ch
itectu
r
e.
I
t f
o
r
m
s
a
web
o
f
c
o
n
n
ec
tio
n
s
b
y
lin
k
in
g
ea
c
h
n
eu
r
o
n
in
th
e
c
u
r
r
en
t la
y
er
t
o
ev
er
y
n
eu
r
o
n
in
t
h
e
p
r
ev
io
u
s
la
y
e
r
,
r
esu
ltin
g
i
n
a
f
u
lly
in
te
r
co
n
n
ec
ted
s
tr
u
ctu
r
e.
T
h
is
d
esig
n
en
s
u
r
es
th
at
in
f
o
r
m
atio
n
f
lo
ws s
ea
m
less
ly
ac
r
o
s
s
th
e
n
etwo
r
k
,
f
ac
ilit
atin
g
co
m
p
lex
p
atter
n
r
ec
o
g
n
itio
n
.
So
f
tMa
x
lay
er
: th
e
o
u
tp
u
t
o
f
th
e
f
u
lly
co
n
n
ec
ted
l
ay
er
is
to
b
e
ca
teg
o
r
ized
b
y
th
i
s
lay
er
.
C
NN:
d
ee
p
n
eu
r
al
n
etwo
r
k
s
with
a
k
n
o
wn
g
r
id
-
lik
e
t
o
p
o
lo
g
y
ar
e
k
n
o
wn
as
C
NNs.
C
NNs
ar
e
d
esig
n
ed
f
o
r
d
ata
with
th
is
ty
p
e
o
f
to
p
o
lo
g
y
.
T
h
ese
n
etwo
r
k
s
u
s
e
th
e
co
n
v
o
lu
tio
n
o
p
er
at
io
n
,
as
th
eir
n
a
m
e
s
u
g
g
ests
.
Du
e
to
th
eir
ca
p
ac
ity
to
o
cc
u
p
y
th
e
t
o
p
o
lo
g
y
o
f
illu
s
tr
atio
n
s
,
C
NNs
is
wi
d
ely
u
s
ed
in
im
ag
e
r
ec
o
g
n
itio
n
.
T
h
e
n
,
as
a
r
esu
lt
o
f
its
im
p
r
o
v
em
en
ts
with
im
a
g
es,
C
NNs
is
u
s
ed
in
o
th
er
d
o
m
ain
s
,
s
u
ch
as
h
a
n
d
g
estu
r
e
id
en
tific
atio
n
an
d
HAR.
T
h
e
C
NN
ar
ch
itectu
r
e
co
n
s
is
ts
o
f
m
u
ltip
le
lay
er
s
:
in
p
u
t,
co
n
v
o
lu
tio
n
,
p
u
llin
g
,
o
u
t
p
u
t,
an
d
f
u
lly
co
n
n
ec
ted
.
Fig
u
r
e
3
d
e
p
icts
a
C
NN
ar
ch
itectu
r
e
co
n
s
is
tin
g
o
f
an
in
p
u
t
lay
er
,
co
n
v
o
l
u
tio
n
,
a
p
o
o
lin
g
la
y
er
,
two
f
u
lly
lin
k
ed
lay
e
r
s
,
an
d
a
n
o
u
tp
u
t
lay
er
.
T
h
e
co
n
v
o
lu
ti
o
n
lay
er
g
en
e
r
ates
f
ea
t
u
r
e
m
ap
s
f
r
o
m
in
p
u
t
d
ata
o
r
o
u
tp
u
t
f
r
o
m
th
e
p
r
i
o
r
lev
el
b
y
m
u
ltip
ly
in
g
f
ilter
s
,
in
p
u
t
d
ata,
o
r
o
u
tp
u
t
f
r
o
m
th
e
p
r
ec
ed
in
g
lay
e
r
elem
en
t
-
by
-
elem
e
n
t.
E
ac
h
f
ea
tu
r
e
m
ap
is
s
u
b
jecte
d
to
a
p
o
o
lin
g
lay
er
af
ter
th
e
co
n
v
o
l
u
tio
n
lay
er
,
wh
ich
r
ed
u
ce
s
th
e
am
o
u
n
t
o
f
C
NN
co
m
p
u
tatio
n
n
ee
d
e
d
b
y
d
o
wn
s
ca
lin
g
th
e
s
p
atial
s
ize.
All o
f
th
e
n
o
d
es in
th
e
f
u
lly
co
n
n
ec
ted
lay
er
s
ar
e
co
n
n
ec
ted
t
o
ev
er
y
o
th
er
n
o
d
e
in
th
e
lay
e
r
b
elo
w,
ju
s
t lik
e
in
FF
NN.
T
h
e
o
u
tp
u
t
lay
er
s
u
b
s
eq
u
en
tly
u
s
es
ac
tiv
atio
n
f
u
n
ct
io
n
s
to
g
et
th
e
o
u
tp
u
ts
;
“
So
f
t
Ma
x
”
is
a
co
m
m
o
n
ac
tiv
atio
n
f
u
n
ctio
n
f
o
r
class
if
icatio
n
p
r
o
b
lem
s
.
T
h
e
weig
h
ts
in
th
e
f
u
lly
co
n
n
ec
ted
lay
e
r
s
a
n
d
lear
n
a
b
le
f
ilter
s
in
th
e
c
o
n
v
o
lu
tio
n
al
lay
er
s
ar
e
c
h
an
g
e
d
u
s
in
g
o
p
tim
izatio
n
tech
n
i
q
u
es
lik
e
g
r
a
d
ien
t
d
escen
t
a
n
d
a
b
ac
k
p
r
o
p
ag
atio
n
ap
p
r
o
ac
h
o
n
c
e
th
e
er
r
o
r
h
as b
ee
n
ca
lcu
lated
.
Fig
u
r
e
3
.
C
NN
ar
ch
itectu
r
e
T
h
e
f
u
n
d
am
en
tal
L
STM
u
n
it
,
wh
ich
co
n
s
is
ts
o
f
a
ce
ll
wi
th
an
in
p
u
t,
o
u
tp
u
t,
a
n
d
f
o
r
g
et
g
ate,
is
s
h
o
wn
in
Fig
u
r
e
4
.
T
h
e
c
o
n
ce
p
t
o
f
g
atin
g
is
u
s
ed
b
y
L
STM
s
to
ad
d
r
ess
th
e
ex
p
lo
d
in
g
g
r
ad
ien
t
p
r
o
b
lem
.
E
ac
h
o
f
th
e
th
r
ee
g
ates
ca
n
b
e
th
o
u
g
h
t
o
f
as
a
tr
ad
itio
n
al
a
r
tific
ia
l
n
eu
r
o
n
,
ca
lcu
latin
g
a
n
ac
tiv
a
tio
n
o
f
a
weig
h
ted
s
u
m
o
f
th
e
“
h
i
d
d
en
s
tate
”
g
_
(
j
-
1
)
f
r
o
m
th
e
p
r
e
v
io
u
s
tim
e
s
tep
,
an
y
b
ias
b
i,
an
d
th
e
cu
r
r
e
n
t
d
ata
k
_
j.
R
em
em
b
er
in
g
v
alu
es
ac
r
o
s
s
u
n
p
r
ed
ictab
le
tim
e
p
e
r
io
d
s
is
th
e
ce
ll'
s
r
esp
o
n
s
ib
ilit
y
.
T
h
e
f
lo
w
o
f
v
alu
es
ac
r
o
s
s
th
e
L
STM
co
n
n
ec
tio
n
s
ca
n
b
e
th
o
u
g
h
t
o
f
as
b
ein
g
c
o
n
tr
o
lle
d
b
y
th
e
g
ates.
At
ea
ch
tim
e
s
tep
,
th
ey
d
eter
m
in
e
wh
ich
o
f
th
e
f
o
llo
win
g
o
p
er
at
io
n
s
th
e
ce
ll
will
p
er
f
o
r
m
.
is
t
h
e
weig
h
ts
co
n
n
ec
ted
to
ea
c
h
m
u
ltip
licatio
n
at
g
ate
in
(
7
)
to
(
1
2
)
,
an
d
ar
e
p
o
ten
tial
ch
o
ices
f
o
r
ac
tiv
atio
n
f
u
n
ctio
n
s
.
T
h
e
in
p
u
t
g
ate
in
Fig
u
r
e
4
r
eg
u
lates h
o
w
m
u
ch
o
f
a
u
n
f
a
m
iliar
v
alu
e
f
lo
w
in
t
o
th
e
ce
ll
d
u
r
in
g
a
wr
ite
o
p
e
r
atio
n
in
(
7
)
.
=
(
[
−
1
,
]
+
)
(
7
)
Similar
wo
r
k
is
d
o
n
e
b
y
th
e
f
o
r
g
et
g
at
e,
wh
ich
r
eg
u
lates
h
o
w
m
u
ch
o
f
th
e
c
u
r
r
e
n
t
ce
ll
v
alu
e
is
r
etain
ed
wh
ile
p
er
f
o
r
m
i
n
g
a
r
e
s
et:
=
(
[
−
1
,
]
)
(
8
)
T
h
e
p
o
ten
tial m
e
m
o
r
y
ce
ll is
u
p
d
ated
in
a
m
an
n
e
r
s
im
ilar
to
(
9
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
R
ev
o
lu
tio
n
iz
in
g
h
u
ma
n
a
ctivity
r
ec
o
g
n
itio
n
w
ith
p
r
o
p
h
et
a
lg
o
r
ith
m
a
n
d
d
ee
p
… (
J
a
yk
u
ma
r
S
.
Dh
a
g
e
)
1113
̃
=
(
[
−
1
,
]
+
)
(
9
)
A
n
d
th
e
in
s
id
e
lo
n
g
-
ter
m
m
e
m
o
r
y
,
also
k
n
o
wn
as
th
e
f
u
r
th
er
ce
ll
m
em
o
r
y
,
is
p
r
o
d
u
ce
d
b
y
co
m
b
in
in
g
th
ese
v
ar
io
u
s
in
ter
n
al
v
alu
es a
s
(
1
0
)
.
ℎ
=
∘
∘
̃
(
1
0
)
As
a
r
esu
lt,
th
e
o
u
tp
u
t
g
ate
g
en
er
ates
th
e
ce
ll
o
u
tp
u
t
to
r
eg
u
late
th
e
d
eg
r
ee
to
wh
ich
th
e
v
alu
e
in
th
e
ce
ll
is
u
s
ed
to
wo
r
k
o
u
t th
e
o
u
tp
u
t a
c
tiv
atio
n
,
p
er
f
o
r
m
in
g
a
r
ea
d
o
p
er
atio
n
is
ex
p
r
ess
ed
in
(
1
1
)
.
ℎ
=
(
[
−
1
,
]
ℎ
)
(
1
1
)
T
h
e
h
id
d
e
n
ce
ll'
s
o
u
tp
u
t
is
ev
en
tu
ally
d
is
co
v
e
r
ed
a
n
d
u
s
e
d
to
co
m
m
u
n
icate
with
else
ce
lls
in
th
e
d
ee
p
n
etwo
r
k
.
A
d
ee
p
n
etw
o
r
k
with
m
an
y
u
n
its
p
r
esen
t
h
as
a
lo
t
o
f
p
ar
am
eter
s
b
ec
au
s
e
ea
ch
g
ate
h
as
p
ar
am
eter
s
f
o
r
its
weig
h
ts
an
d
b
iases
.
T
h
r
o
u
g
h
n
etwo
r
k
tr
ai
n
in
g
,
th
e
weig
h
ts
o
f
th
ese
ass
o
ciatio
n
s
ar
e
eith
er
k
n
o
wled
g
ea
b
le
o
r
u
p
-
to
-
d
ate.
T
h
e
m
ath
em
atica
l m
o
d
el
is
s
h
o
wn
in
(
1
2
)
=
ℎ
∘
(
)
(
1
2
)
Fig
u
r
e
4
.
Ar
c
h
itectu
r
e
o
f
L
ST
M
4.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
e
r
ea
l
tim
e
d
ata
s
et
is
cr
ea
ted
an
d
n
am
ed
as
s
en
s
o
r
d
ata
s
et.
T
h
er
e
ar
e
f
o
llo
win
g
ac
tiv
ity
lab
els
u
s
ed
f
o
r
th
e
p
r
ed
ictio
n
a
n
d
an
aly
s
is
.
An
ap
p
is
d
ev
elo
p
ed
to
co
llect
d
ata
f
r
o
m
th
e
p
eo
p
le.
6
ac
tiv
ity
lab
els ar
e:
0
-
lay
in
g
,
1
-
s
itti
n
g
,
2
-
s
tan
d
in
g
,
3
-
walk
in
g
,
4
-
walk
in
g
d
o
w
n
s
tair
s
5
-
walk
in
g
u
p
s
tair
s
.
T
h
e
d
ata
s
et
h
as
a
to
tal
1
,
1
0
7
r
ec
o
r
d
s
.
Her
e
f
o
r
th
e
tr
ai
n
in
g
p
u
r
p
o
s
e,
7
0
% o
f
d
ata
i.e
.
7
7
4
r
ec
o
r
d
s
ar
e
u
s
ed
wh
er
ea
s
3
0
%
o
f
d
ata
i.e
.
3
3
3
r
e
co
r
d
s
ar
e
u
s
ed
f
o
r
t
h
e
test
in
g
p
u
r
p
o
s
e.
Fig
u
r
e
5
d
is
p
lay
th
e
co
n
f
u
s
io
n
m
atr
ix
o
f
d
if
f
er
e
n
t
m
o
d
els.
F
r
o
m
Fig
u
r
e
5
(
a)
,
it
ca
n
b
e
o
b
s
er
v
ed
t
h
at
th
e
L
R
m
o
d
el
g
iv
es
least
ac
cu
r
ate
r
esu
lts
f
o
r
“
STAN
DI
NG
”
lab
el
as
co
m
p
a
r
ed
to
o
th
er
5
la
b
els
an
d
th
e
lab
el
“
W
AL
KI
NG
”
g
iv
es
th
e
m
o
s
t
ac
cu
r
ate
r
esu
lt.
T
h
e
ac
c
u
r
ac
y
is
9
7
.
3
9
%.
I
n
Fig
u
r
e
5
(
b
)
,
it
ca
n
b
e
o
b
s
er
v
ed
th
at
th
e
SVM
m
o
d
el
g
iv
es
th
e
least
ac
cu
r
ate
r
esu
lt
s
f
o
r
“
STAN
DI
NG
”
lab
el
as
co
m
p
ar
ed
to
o
th
er
5
lab
els
an
d
th
e
la
b
el
“
W
AL
KI
NG
”
g
iv
es
th
e
m
o
s
t
ac
cu
r
ate
r
esu
lt.
T
h
e
ac
cu
r
ac
y
is
9
7
.
7
9
%.
I
n
Fig
u
r
e
5
(
c)
,
it
ca
n
b
e
o
b
s
er
v
ed
t
h
at
t
h
e
DT
m
o
d
el
g
iv
es
th
e
least
ac
cu
r
ate
r
esu
lts
f
o
r
“
W
AL
KI
NG
DO
W
NST
AI
R
S
”
lab
el
as
co
m
p
ar
e
d
to
o
th
er
5
lab
els
an
d
th
e
lab
el
“
W
AL
KI
NG
”
g
iv
es
th
e
m
o
s
t
ac
cu
r
ate
r
esu
lt.
T
h
e
ac
cu
r
ac
y
is
9
2
.
7
9
%.
I
n
Fig
u
r
e
5
(
d
)
,
it
ca
n
b
e
o
b
s
er
v
ed
th
a
t
th
e
R
F
m
o
d
el
g
iv
es
th
e
lea
s
t
ac
cu
r
ate
r
esu
lts
f
o
r
“
STAN
DI
NG
”
lab
el
as
co
m
p
ar
ed
to
o
th
er
5
lab
els
an
d
th
e
lab
el
“
W
AL
KI
NG
”
g
iv
es
th
e
m
o
s
t
ac
cu
r
ate
r
esu
lts
.
T
h
e
ac
cu
r
ac
y
is
9
6
.
1
9
%.
I
n
Fig
u
r
e
5
(
e)
,
it
ca
n
b
e
o
b
s
er
v
ed
th
at
th
e
NN
g
iv
es
least
ac
cu
r
ate
r
esu
lts
f
o
r
th
e
“
W
AL
KI
NG
DO
W
NSTA
I
R
S
”
la
b
el
as
co
m
p
ar
ed
to
r
em
ain
in
g
5
lab
els
an
d
th
e
lab
el
“
W
AL
KI
NG
”
g
iv
es
th
e
m
o
s
t
ac
cu
r
ate
r
esu
lts
.
T
h
e
ac
cu
r
ac
y
is
9
7
.
6
9
%.
T
h
e
o
th
er
p
er
f
o
r
m
an
ce
p
ar
a
m
eter
s
p
r
ec
is
i
o
n
,
F1
s
co
r
e
,
a
n
d
r
ec
a
ll
ar
e
also
ta
k
en
i
n
to
c
o
n
s
id
er
atio
n
.
As
s
h
o
wn
in
Fig
u
r
e
5
(
f
)
a
n
d
5
(
g
)
,
w
h
en
c
o
n
v
o
lu
ti
o
n
is
a
p
p
lied
,
th
e
m
o
s
t
ac
cu
r
ate
r
esu
lts
ar
e
o
b
tain
e
d
.
T
h
e
co
n
v
o
lu
tio
n
g
iv
es a
cc
u
r
ac
y
as 9
8
.
4
9
% a
n
d
L
STM
with
co
n
v
o
lu
tio
n
g
iv
e
n
ac
cu
r
ac
y
as 9
8
.
0
9
%.
All Fig
u
r
es a
r
e
m
o
d
if
ie
d
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
7
7
6
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
,
Vo
l.
14
,
No
.
3
,
Dec
em
b
er
20
25
:
1
1
0
8
-
1
1
1
8
1114
(
a)
(
b
)
(
c)
(
d
)
(
e)
(f)
(
g
)
F
i
g
u
r
e
5
.
C
o
n
f
u
s
i
o
n
m
a
t
r
i
x
o
f
:
(
a
)
L
R
,
(
b
)
S
VM
,
(
c
)
D
T
,
(
d
)
R
F
,
(
e
)
N
N
,
(
f
)
T
D
L
-
L
S
T
M
,
a
n
d
(
g
)
L
S
T
M
-
T
D
L
T
h
e
r
esu
lts
in
T
ab
le
1
m
ak
e
it
clea
r
th
at
ad
v
a
n
ce
d
d
ee
p
lear
n
in
g
tech
n
iq
u
es
h
a
v
e
a
m
ajo
r
im
p
ac
t
o
n
im
p
r
o
v
in
g
th
e
ac
cu
r
ac
y
o
f
HAR.
Fo
r
ex
am
p
le,
th
e
ac
tiv
ity
lab
eled
“
W
AL
KI
NG
”
h
as
b
ee
n
p
r
ed
icted
with
to
p
-
n
o
tc
h
ac
c
u
r
ac
y
,
b
ea
tin
g
o
u
t
o
ld
e
r
,
m
o
r
e
tr
ad
itio
n
al
m
e
t
h
o
d
s
.
T
h
is
s
h
o
ws
ju
s
t
h
o
w
p
o
wer
f
u
l
d
ee
p
lea
r
n
in
g
ca
n
b
e
wh
en
it
co
m
es
to
wo
r
k
in
g
with
co
m
p
lex
d
atasets
an
d
ac
cu
r
ately
p
r
ed
ictin
g
ac
ti
v
ities
.
T
h
e
o
v
er
all
m
o
d
el
ac
c
u
r
ac
y
s
aw
a
n
o
ticea
b
le
ju
m
p
to
9
8
.
4
9
%
an
d
9
8
.
0
9
%
with
th
e
latest
d
ee
p
lear
n
in
g
m
o
d
els,
wh
ich
r
ea
lly
u
n
d
e
r
s
co
r
es th
e
p
o
ten
tial th
ese
tech
n
iq
u
es h
av
e
to
elev
ate
HAR s
y
s
tem
s
.
F
i
g
u
r
e
6
s
h
o
w
s
t
h
e
p
e
r
f
o
r
m
a
n
c
e
o
f
v
a
r
i
o
u
s
M
L
m
o
d
e
l
s
i
n
H
A
R
,
s
h
o
wi
n
g
m
e
t
r
i
c
s
s
u
c
h
a
s
F
i
g
u
r
e
6
(
a
)
ac
cu
r
ac
y
,
Fig
u
r
e
6
(
b
)
p
r
ec
is
io
n
,
Fig
u
r
e
6
(
c)
r
ec
all,
an
d
Fig
u
r
e
6
(
d
)
F1
s
co
r
e
.
Ad
v
a
n
ce
d
d
ee
p
lear
n
in
g
m
o
d
els
-
T
DL
-
L
STM
an
d
L
STM
-
T
DL
-
o
u
tp
er
f
o
r
m
tr
ad
itio
n
a
l
m
o
d
els
lik
e
L
R
,
SVM,
DT
,
R
F,
an
d
NN,
with
th
e
h
ig
h
est
ac
cu
r
ac
y
o
f
9
8
.
4
9
%
an
d
9
8
.
0
9
%,
r
esp
ec
tiv
ely
.
SV
M
an
d
NN
also
p
er
f
o
r
m
well
with
ac
cu
r
ac
y
ar
o
u
n
d
9
7
.
7
9
%,
wh
ile
DT
an
d
R
F
lag
,
esp
ec
ially
in
ac
cu
r
ac
y
(
9
2
.
7
9
%
a
n
d
9
6
.
6
9
%,
r
e
s
p
ec
tiv
ely
)
.
T
h
es
e
r
esu
lts
h
ig
h
lig
h
t
th
e
s
u
p
er
io
r
p
er
f
o
r
m
an
ce
o
f
d
ee
p
lear
n
in
g
ap
p
r
o
ac
h
es
f
o
r
HAR,
p
ar
ti
cu
lar
ly
in
h
an
d
lin
g
co
m
p
lex
d
a
ta
an
d
d
eliv
er
in
g
h
ig
h
er
ac
cu
r
ac
y
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
R
ev
o
lu
tio
n
iz
in
g
h
u
ma
n
a
ctivity
r
ec
o
g
n
itio
n
w
ith
p
r
o
p
h
et
a
lg
o
r
ith
m
a
n
d
d
ee
p
… (
J
a
yk
u
ma
r
S
.
Dh
a
g
e
)
1115
T
ab
le
1
.
Statis
tics
o
f
v
ar
io
u
s
a
lg
o
r
ith
m
s
S
r
.
N
o
.
A
l
g
o
r
i
t
h
m
A
c
c
u
r
a
c
y
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F
1
s
c
o
r
e
1
LR
9
7
.
3
9
0
.
9
5
6
0
.
9
4
0
.
9
4
2
S
V
M
9
7
.
7
9
0
.
9
6
0
.
9
5
0
.
9
5
3
DT
9
2
.
7
9
0
.
9
6
0
.
8
5
0
.
8
5
4
RF
9
6
.
1
9
0
.
9
2
0
.
9
1
0
.
9
1
5
NN
9
7
.
6
9
0
.
9
4
0
.
9
8
0
.
9
9
6
TD
L
-
LSTM
9
8
.
4
9
0
.
9
5
0
.
9
5
0
.
9
6
7
LSTM
-
TD
L
9
8
.
0
9
0
.
9
8
0
.
9
8
0
.
9
8
(
a)
(
b
)
(
c)
(
d
)
Fig
u
r
e
6
.
Gr
a
p
h
ical
r
ep
r
esen
tatio
n
o
f
:
(
a
)
a
cc
u
r
ac
y
,
(
b
)
p
r
ec
is
io
n
,
(
c)
r
ec
all,
a
n
d
(
d
)
F1
s
co
r
e
Ou
r
s
tu
d
y
s
u
g
g
ests
th
at
th
e
p
r
o
p
o
s
ed
T
DL
-
L
STM
an
d
L
STM
-
T
DL
m
o
d
els
ac
h
iev
e
h
ig
h
e
r
ac
cu
r
ac
y
in
HAR
co
m
p
ar
ed
to
tr
ad
itio
n
al
ML
m
o
d
els
an
d
ea
r
lier
d
ee
p
lear
n
in
g
ap
p
r
o
ac
h
es.
W
h
ile
Mu
r
alid
h
ar
an
et
a
l.
[
2
1
]
r
ep
o
r
ted
a
v
alid
atio
n
ac
c
u
r
ac
y
o
f
9
6
.
1
3
%
u
s
in
g
a
1
-
D
C
NN,
o
u
r
m
o
d
els
s
ig
n
if
ican
t
ly
o
u
tp
e
r
f
o
r
m
th
is
with
ac
cu
r
ac
y
o
f
9
8
.
4
9
%.
S
im
ilar
ly
,
Kir
an
y
az
et
a
l.
[
2
3
]
h
ig
h
lig
h
ted
th
e
ef
f
icien
cy
o
f
1
-
D
C
NNs
in
en
g
in
ee
r
in
g
ap
p
licatio
n
s
with
m
in
im
al
co
m
p
u
tatio
n
al
co
m
p
l
ex
ity
,
y
et
th
eir
s
co
p
e
was lim
i
ted
to
th
r
ee
ac
tiv
ity
lab
els.
I
n
co
n
tr
ast,
o
u
r
m
o
d
els
wer
e
test
ed
o
n
a
co
m
p
r
eh
en
s
i
v
e
d
ataset
with
s
ix
ac
tiv
ity
lab
els,
d
em
o
n
s
tr
atin
g
s
u
p
er
io
r
ac
c
u
r
ac
y
an
d
r
o
b
u
s
tn
ess
.
C
o
m
p
ar
in
g
with
Min
ar
n
o
et
a
l.
[
2
4
]
,
wh
o
ac
h
iev
ed
a
m
a
x
im
u
m
ac
c
u
r
ac
y
o
f
9
8
.
9
6
%
u
s
in
g
SVM
with
an
R
B
F
k
er
n
el,
o
u
r
m
o
d
els
p
r
o
v
id
e
co
m
p
ar
ab
le
p
er
f
o
r
m
an
c
e
wh
ile
ad
d
r
ess
in
g
tem
p
o
r
al
d
ep
en
d
en
cies
m
o
r
e
ef
f
ec
tiv
ely
th
r
o
u
g
h
L
STM
ar
ch
itectu
r
e.
A
d
d
itio
n
ally
,
u
n
lik
e
ap
p
r
o
ac
h
es
b
y
Z
h
an
g
et
a
l.
[
2
5
]
a
n
d
L
ee
e
t
a
l.
[
4
]
,
wh
i
ch
f
o
cu
s
ed
o
n
co
n
v
o
l
u
tio
n
al
n
etwo
r
k
s
ac
h
ie
v
in
g
ac
cu
r
ac
ies
o
f
9
2
.
7
1
%
an
d
b
elo
w,
o
u
r
m
o
d
el
s
in
co
r
p
o
r
ate
b
o
th
co
n
v
o
lu
tio
n
an
d
r
ec
u
r
r
en
t
lay
e
r
s
,
en
s
u
r
i
n
g
b
etter
ca
p
tu
r
e
o
f
s
eq
u
en
tial
d
ata
p
atter
n
s
.
T
h
e
r
esu
lts
s
u
g
g
est
th
at
in
co
r
p
o
r
ati
n
g
ad
v
a
n
ce
d
d
ee
p
lear
n
i
n
g
ar
ch
itectu
r
es
s
u
ch
as
T
DL
-
L
STM
an
d
L
STM
-
T
DL
m
ay
b
e
n
ef
it
HAR
task
s
b
y
en
h
an
cin
g
ac
cu
r
ac
y
an
d
r
e
d
u
cin
g
f
alse
p
o
s
itiv
es
with
o
u
t
ad
v
er
s
ely
im
p
ac
tin
g
co
m
p
u
tatio
n
al
ef
f
icien
cy
.
T
h
e
s
e
f
in
d
in
g
s
h
ig
h
lig
h
t
th
e
tr
an
s
f
o
r
m
ativ
e
p
o
ten
tial
o
f
d
ee
p
lear
n
in
g
in
ad
v
an
cin
g
H
AR
b
ey
o
n
d
tr
ad
itio
n
al
ap
p
r
o
ac
h
es.
T
h
is
s
tu
d
y
ex
p
lo
r
ed
a
co
m
p
r
e
h
en
s
iv
e
d
ataset
with
s
ix
ac
tiv
ity
lab
els
u
s
in
g
ad
v
an
ce
d
d
ee
p
lear
n
in
g
m
o
d
els,
T
DL
-
L
STM
an
d
L
STM
-
T
DL
,
ac
h
iev
in
g
h
i
g
h
ac
c
u
r
ac
y
an
d
r
o
b
u
s
t
p
er
f
o
r
m
a
n
c
e.
Ho
wev
er
,
f
u
r
th
e
r
an
d
in
-
d
e
p
th
s
tu
d
ies m
ay
b
e
n
ee
d
ed
to
co
n
f
ir
m
th
e
g
en
er
aliz
ab
ilit
y
o
f
th
ese
m
o
d
els,
esp
ec
ially
r
eg
ar
d
in
g
th
eir
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
7
7
6
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
,
Vo
l.
14
,
No
.
3
,
Dec
em
b
er
20
25
:
1
1
0
8
-
1
1
1
8
1116
ap
p
licatio
n
in
r
ea
l
-
wo
r
l
d
en
v
i
r
o
n
m
en
ts
with
d
iv
er
s
e
s
en
s
o
r
s
etu
p
s
an
d
v
ar
ied
p
o
p
u
latio
n
d
em
o
g
r
a
p
h
ics.
T
h
e
d
ataset
u
s
ed
,
wh
ile
ex
ten
s
iv
e,
was
lim
ited
to
c
o
n
tr
o
lled
co
n
d
itio
n
s
an
d
m
ay
n
o
t f
u
lly
ca
p
tu
r
e
th
e
co
m
p
lex
ities
o
f
n
atu
r
al
ac
tiv
ity
p
atter
n
s
.
Ad
d
itio
n
ally
,
th
e
co
m
p
u
tatio
n
al
ef
f
icien
cy
o
f
th
e
p
r
o
p
o
s
ed
m
o
d
els
in
r
eso
u
r
ce
-
co
n
s
tr
ain
ed
d
e
v
ices lik
e
s
m
ar
tp
h
o
n
es n
ee
d
s
f
u
r
th
e
r
ev
alu
atio
n
to
en
s
u
r
e
p
r
ac
tical
d
ep
l
o
y
m
e
n
t.
5.
CO
NCLU
S
I
O
N
AND
F
U
T
U
RE
WO
RK
R
ec
en
t
o
b
s
er
v
atio
n
s
s
u
g
g
est
th
at
ad
v
a
n
ce
d
d
ee
p
lear
n
i
n
g
m
o
d
els
s
ig
n
if
ican
tly
en
h
a
n
ce
HAR
ac
cu
r
ac
y
.
O
u
r
f
i
n
d
in
g
s
p
r
o
v
id
e
co
n
clu
s
iv
e
e
v
id
en
ce
t
h
at
th
i
s
im
p
r
o
v
em
e
n
t
is
d
r
iv
e
n
b
y
t
h
e
ef
f
ec
tiv
e
n
ess
o
f
alg
o
r
ith
m
s
lik
e
T
DL
-
L
STM
an
d
L
STM
-
T
DL
,
w
h
ich
ac
h
iev
e
im
p
r
ess
iv
e
ac
cu
r
ac
ies
o
f
9
8
.
4
9
%
an
d
9
8
.
0
9
%
,
r
esp
ec
tiv
ely
,
s
u
r
p
ass
in
g
tr
ad
it
io
n
al
ML
m
eth
o
d
s
.
T
h
is
ad
v
an
ce
m
en
t
is
attr
ib
u
ted
to
th
e
m
o
d
els'
ab
ilit
y
to
h
an
d
le
lar
g
e
d
atasets
an
d
ca
p
tu
r
e
s
u
b
tle
p
atter
n
s
.
T
h
e
p
r
esen
t
wo
r
k
h
ig
h
lig
h
ts
o
p
p
o
r
t
u
n
ities
f
o
r
g
r
o
wth
,
in
clu
d
in
g
ex
p
an
d
i
n
g
d
atasets
,
in
co
r
p
o
r
atin
g
v
ar
ied
ac
tiv
ities
an
d
s
en
s
o
r
s
,
an
d
ad
ap
tin
g
to
r
ea
l
-
wo
r
l
d
co
n
d
itio
n
s
to
en
h
an
ce
ac
cu
r
a
cy
an
d
g
e
n
er
aliza
b
ilit
y
.
L
o
o
k
in
g
ah
ea
d
,
a
d
v
an
ce
m
e
n
ts
in
AI
an
d
ML
co
u
ld
in
tr
o
d
u
ce
n
o
v
el
ar
c
h
itectu
r
es
an
d
alg
o
r
ith
m
s
,
p
u
s
h
in
g
HA
R
s
y
s
tem
s
f
u
r
th
er
an
d
im
p
r
o
v
in
g
ap
p
licatio
n
s
in
h
ea
lth
ca
r
e,
s
ec
u
r
ity
,
a
n
d
s
m
a
r
t
h
o
m
es.
T
h
is
s
tu
d
y
u
n
d
er
s
c
o
r
es
th
e
tr
an
s
f
o
r
m
ativ
e
r
o
le
o
f
d
ee
p
lear
n
in
g
in
HAR an
d
lay
s
th
e
g
r
o
u
n
d
wo
r
k
f
o
r
f
u
tu
r
e
i
n
n
o
v
atio
n
.
F
UT
UR
E
WO
RK
Fu
tu
r
e
r
esear
ch
ca
n
b
u
ild
u
p
o
n
th
is
s
tu
d
y
b
y
f
o
cu
s
in
g
o
n
ex
p
an
d
in
g
d
atasets
to
in
clu
d
e
m
o
r
e
d
iv
er
s
e
ac
tiv
ities
,
in
d
iv
id
u
als,
an
d
r
ea
l
-
wo
r
ld
co
n
d
itio
n
s
,
en
h
an
c
in
g
th
e
g
en
er
aliza
b
ilit
y
an
d
r
o
b
u
s
tn
ess
o
f
th
e
m
o
d
els.
I
n
co
r
p
o
r
atin
g
ad
d
itio
n
al
s
en
s
o
r
ty
p
es
an
d
ex
p
l
o
r
in
g
m
u
lti
-
m
o
d
al
a
p
p
r
o
ac
h
es
co
u
ld
f
u
r
th
e
r
im
p
r
o
v
e
ac
tiv
ity
r
ec
o
g
n
itio
n
p
er
f
o
r
m
a
n
ce
.
Ad
d
itio
n
ally
,
i
n
v
esti
g
atin
g
n
o
v
el
d
ee
p
lear
n
in
g
a
r
ch
i
tectu
r
es
o
r
h
y
b
r
id
m
o
d
els
m
ig
h
t
p
u
s
h
t
h
e
b
o
u
n
d
ar
ies
o
f
ac
cu
r
ac
y
a
n
d
ad
ap
tab
i
lity
in
HAR
s
y
s
tem
s
.
Ad
d
r
ess
in
g
c
h
allen
g
es
lik
e
r
ea
l
-
tim
e
p
r
o
ce
s
s
in
g
an
d
c
o
m
p
u
tatio
n
al
ef
f
icien
cy
will
also
b
e
p
iv
o
tal
in
b
r
o
ad
e
n
in
g
th
e
p
r
ac
tical
ap
p
licatio
n
s
o
f
th
ese
s
y
s
tem
s
ac
r
o
s
s
d
o
m
ain
s
s
u
ch
as h
ea
lth
ca
r
e,
s
ec
u
r
ity
,
an
d
s
m
ar
t e
n
v
ir
o
n
m
en
ts
.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
Au
th
o
r
s
s
tate
th
er
e
is
n
o
f
u
n
d
i
n
g
in
v
o
lv
ed
.
AUTHO
R
CO
NT
RI
B
UT
I
O
NS ST
A
T
E
M
E
N
T
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
J
ay
k
u
m
ar
S.
Dh
a
g
e
✓
✓
✓
✓
✓
✓
✓
✓
✓
Av
in
ash
K.
Gu
lv
e
✓
✓
✓
✓
✓
✓
✓
✓
✓
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
th
at
th
e
y
h
av
e
n
o
co
n
f
lict o
f
in
ter
est.
I
NF
O
RM
E
D
CO
NS
E
N
T
I
n
f
o
r
m
ed
co
n
s
en
t w
as o
b
tain
e
d
f
r
o
m
all
in
d
iv
id
u
al
p
ar
ticip
a
n
ts
in
v
o
lv
ed
i
n
th
e
s
tu
d
y
.
E
T
H
I
CAL AP
P
RO
V
AL
T
h
is
s
tu
d
y
d
id
n
o
t
i
n
v
o
lv
e
an
y
e
x
p
er
im
e
n
ts
with
h
u
m
a
n
p
ar
ticip
an
ts
o
r
a
n
im
als,
an
d
th
er
ef
o
r
e
eth
ical
ap
p
r
o
v
al
was n
o
t r
eq
u
i
r
ed
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
R
ev
o
lu
tio
n
iz
in
g
h
u
ma
n
a
ctivity
r
ec
o
g
n
itio
n
w
ith
p
r
o
p
h
et
a
lg
o
r
ith
m
a
n
d
d
ee
p
… (
J
a
yk
u
ma
r
S
.
Dh
a
g
e
)
1117
DATA AV
AI
L
AB
I
L
I
T
Y
T
h
e
d
ata
th
at
s
u
p
p
o
r
t
th
e
f
i
n
d
in
g
s
o
f
th
is
s
tu
d
y
ar
e
av
aila
b
le
f
r
o
m
th
e
c
o
r
r
esp
o
n
d
in
g
a
u
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
]
K
.
C
h
e
n
,
D
.
Z
h
a
n
g
,
L.
Y
a
o
,
B
.
G
u
o
,
Z.
Y
u
,
a
n
d
Y
.
L
i
u
,
“
D
e
e
p
l
e
a
r
n
i
n
g
f
o
r
se
n
so
r
-
b
a
se
d
h
u
m
a
n
a
c
t
i
v
i
t
y
r
e
c
o
g
n
i
t
i
o
n
:
o
v
e
r
v
i
e
w
,
c
h
a
l
l
e
n
g
e
s,
a
n
d
o
p
p
o
r
t
u
n
i
t
i
e
s
,
”
AC
M
C
o
m
p
u
t
i
n
g
S
u
r
v
e
y
s
,
v
o
l
.
5
4
,
n
o
.
4
,
p
p
.
1
–
4
0
,
2
0
2
1
,
d
o
i
:
1
0
.
1
1
4
5
/
3
4
4
7
7
4
4
.
[
2
]
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
p
.
1
–
3
3
,
2
0
1
4
,
d
o
i
:
1
0
.
1
1
4
5
/
2
4
9
9
6
2
1
.
[
3
]
S
.
T
o
n
t
i
,
B
.
M
a
r
z
o
l
i
n
i
,
a
n
d
M
.
B
u
l
g
h
e
r
o
n
i
,
“
S
m
a
r
t
p
h
o
n
e
-
b
a
s
e
d
p
a
ss
i
v
e
s
e
n
si
n
g
f
o
r
b
e
h
a
v
i
o
r
a
l
a
n
d
p
h
y
s
i
c
a
l
mo
n
i
t
o
r
i
n
g
i
n
f
r
e
e
-
l
i
f
e
c
o
n
d
i
t
i
o
n
s
:
t
e
c
h
n
i
c
a
l
u
s
a
b
i
l
i
t
y
s
t
u
d
y
,
”
J
MIR
Bi
o
m
e
d
i
c
a
l
En
g
i
n
e
e
ri
n
g
,
v
o
l
.
6
,
n
o
.
2
,
p
.
e
1
5
4
1
7
,
2
0
2
1
,
d
o
i
:
1
0
.
2
1
9
6
/
1
5
4
1
7
.
[
4
]
S
.
M
.
Le
e
,
S
.
M
.
Y
o
o
n
,
a
n
d
H
.
C
h
o
,
“
H
u
ma
n
a
c
t
i
v
i
t
y
r
e
c
o
g
n
i
t
i
o
n
f
r
o
m
a
c
c
e
l
e
r
o
m
e
t
e
r
d
a
t
a
u
s
i
n
g
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
,
”
i
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
re
n
c
e
o
n
Bi
g
D
a
t
a
a
n
d
S
m
a
r
t
C
o
m
p
u
t
i
n
g
,
p
p
.
1
3
1
–
1
3
4
,
2
0
1
7
,
d
o
i
:
1
0
.
1
1
0
9
/
B
I
G
C
O
M
P
.
2
0
1
7
.
7
8
8
1
7
2
8
.
[
5
]
D
.
A
n
g
u
i
t
a
,
L
.
O
n
e
t
o
,
X
.
P
a
r
r
a
,
a
n
d
J.
L.
R
e
y
e
s
-
O
r
t
i
z
,
“
En
e
r
g
y
e
f
f
i
c
i
e
n
t
s
ma
r
t
p
h
o
n
e
-
b
a
s
e
d
a
c
t
i
v
i
t
y
r
e
c
o
g
n
i
t
i
o
n
u
si
n
g
f
i
x
e
d
-
p
o
i
n
t
a
r
i
t
h
me
t
i
c
,
”
J
o
u
rn
a
l
o
f
U
n
i
v
e
r
sa
l
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
–
S
p
e
c
i
a
l
I
s
su
e
i
n
A
m
b
i
e
n
t
As
s
i
s
t
e
d
L
i
v
i
n
g
:
H
o
m
e
C
a
re
,
v
o
l
.
1
9
,
n
o
.
9
,
2
0
1
3.
[
6
]
D
.
A
n
g
u
i
t
a
,
L.
O
n
e
t
o
,
X
.
P
a
r
r
a
,
a
n
d
J.
L.
R
e
y
e
s
-
O
r
t
i
z
,
“
A
p
u
b
l
i
c
d
o
ma
i
n
d
a
t
a
se
t
f
o
r
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
si
n
g
smar
t
p
h
o
n
e
s
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
2
1
s
t
Eu
r
o
p
e
a
n
S
y
m
p
o
s
i
u
m
o
n
Art
i
f
i
c
i
a
l
N
e
u
ra
l
N
e
t
w
o
rk
s,
C
o
m
p
u
t
a
t
i
o
n
a
l
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
Ma
c
h
i
n
e
L
e
a
r
n
i
ng
,
2
0
1
3
.
[
7
]
R
.
A
.
V
o
i
c
u
,
C
.
D
o
b
r
e
,
L.
B
a
j
e
n
a
r
u
,
a
n
d
R
.
I
.
C
i
o
b
a
n
u
,
“
H
u
ma
n
p
h
y
si
c
a
l
a
c
t
i
v
i
t
y
r
e
c
o
g
n
i
t
i
o
n
u
s
i
n
g
sm
a
r
t
p
h
o
n
e
s
e
n
s
o
r
s
,
”
S
e
n
s
o
rs
,
v
o
l
.
1
9
,
n
o
.
3
,
p
.
4
5
8
,
2
0
1
9
,
d
o
i
:
1
0
.
3
3
9
0
/
s
1
9
0
3
0
4
5
8
.
[
8
]
D
.
A
n
g
u
i
t
a
,
L.
O
n
e
t
o
,
X
.
P
a
r
r
a
,
a
n
d
J
.
L
.
R
e
y
e
s
-
O
r
t
i
z
,
“
H
u
ma
n
ac
t
i
v
i
t
y
r
e
c
o
g
n
i
t
i
o
n
o
n
smar
t
p
h
o
n
e
s
u
s
i
n
g
a
m
u
l
t
i
c
l
a
ss
h
a
r
d
w
a
r
e
-
f
r
i
e
n
d
l
y
s
u
p
p
o
r
t
v
e
c
t
o
r
m
a
c
h
i
n
e
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
4
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
W
o
rks
h
o
p
o
f
Am
b
i
e
n
t
Ass
i
st
e
d
L
i
v
i
n
g
,
p
p
.
2
1
6
–
2
2
3
,
2
0
1
2
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
6
4
2
-
3
5
3
9
5
-
6
_
3
0
.
[
9
]
J.
L.
R
e
y
e
s
-
O
r
t
i
z
,
X
.
P
a
r
r
a
-
Ll
a
n
a
s,
D
.
A
n
g
u
i
t
a
,
J.
C
a
b
e
s
t
a
n
y
,
a
n
d
A
.
C
a
t
a
l
à
,
“
H
u
m
a
n
a
c
t
i
v
i
t
y
a
n
d
m
o
t
i
o
n
d
i
s
o
r
d
e
r
r
e
c
o
g
n
i
t
i
o
n
:
t
o
w
a
r
d
s
sm
a
r
t
e
r
i
n
t
e
r
a
c
t
i
v
e
c
o
g
n
i
t
i
v
e
e
n
v
i
r
o
n
m
e
n
t
s,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
2
1
s
t
E
u
ro
p
e
a
n
S
y
m
p
o
s
i
u
m
o
n
Art
i
f
i
c
i
a
l
N
e
u
r
a
l
N
e
t
w
o
rks,
C
o
m
p
u
t
a
t
i
o
n
a
l
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
M
a
c
h
i
n
e
L
e
a
rn
i
n
g
,
2
0
1
3
.
[
1
0
]
A
.
M
.
-
D
.
H
a
y
a
t
,
B
.
P
.
B
h
u
y
a
n
,
a
n
d
R
.
To
m
a
r
,
“
H
u
m
a
n
a
c
t
i
v
i
t
y
r
e
c
o
g
n
i
t
i
o
n
f
o
r
e
l
d
e
r
l
y
p
e
o
p
l
e
u
s
i
n
g
ma
c
h
i
n
e
a
n
d
d
e
e
p
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
e
s
,
”
I
n
f
o
rm
a
t
i
o
n
Pr
o
c
e
ssi
n
g
a
n
d
Ma
n
a
g
e
m
e
n
t
,
v
o
l
.
1
3
,
p
.
2
7
5
,
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
i
n
f
o
1
3
0
6
0
2
7
5
.
[
1
1
]
S
.
P
a
r
a
sc
h
i
a
k
o
s
,
R
.
C
a
c
h
u
c
h
o
,
M
.
M
o
e
d
,
C
.
V
a
n
H
e
e
ms
t
,
S
.
M
o
o
i
j
a
a
r
t
,
E.
P
.
S
l
a
g
b
o
o
m
,
a
n
d
M
.
B
e
e
k
ma
n
,
“
A
c
t
i
v
i
t
y
r
e
c
o
g
n
i
t
i
o
n
u
si
n
g
w
e
a
r
a
b
l
e
s
e
n
so
r
s
f
o
r
t
r
a
c
k
i
n
g
t
h
e
e
l
d
e
r
l
y
,
”
U
s
e
r M
o
d
e
l
i
n
g
a
n
d
U
ser
-
Ad
a
p
t
e
d
I
n
t
e
r
a
c
t
i
o
n
,
v
o
l
.
3
0
,
n
o
.
3
,
p
p
.
5
6
7
–
6
0
5
,
2
0
2
0
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
1
2
5
7
-
020
-
0
9
2
6
8
-
2.
[
1
2
]
L.
B
.
M
a
r
i
n
h
o
,
A
.
H
.
D
e
S
o
u
z
a
J
ú
n
i
o
r
,
a
n
d
P
.
P
.
R
e
b
o
u
ç
a
s
F
i
l
h
o
,
“
A
n
e
w
a
p
p
r
o
a
c
h
t
o
h
u
ma
n
a
c
t
i
v
i
t
y
r
e
c
o
g
n
i
t
i
o
n
u
si
n
g
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
t
e
c
h
n
i
q
u
e
s
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
I
n
t
e
l
l
i
g
e
n
t
S
y
st
e
m
s
D
e
si
g
n
a
n
d
A
p
p
l
i
c
a
t
i
o
n
s
,
p
p
.
5
2
9
–
5
3
8
,
2
0
1
6
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
3
1
9
-
5
3
4
8
0
-
0
_
5
2
.
[
1
3
]
I
.
A
.
T.
H
a
sh
e
m,
I
.
Y
a
q
o
o
b
,
N
.
B
.
A
n
u
a
r
,
S
.
M
o
k
h
t
a
r
,
A
.
G
a
n
i
,
a
n
d
S
.
U
.
K
h
a
n
,
“
T
h
e
r
i
se
o
f
'b
i
g
d
a
t
a
'
o
n
c
l
o
u
d
c
o
mp
u
t
i
n
g
:
r
e
v
i
e
w
a
n
d
o
p
e
n
r
e
se
a
r
c
h
i
ssu
e
s
,
”
I
n
f
o
rm
a
t
i
o
n
S
y
s
t
e
m
s
,
v
o
l
.
4
7
,
p
p
.
9
8
–
1
1
5
,
2
0
1
5
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
s.
2
0
1
4
.
0
7
.
0
0
6
.
[
1
4
]
J.
R
.
K
w
a
p
i
s
z
,
G
.
M
.
W
e
i
ss,
a
n
d
S
.
A
.
M
o
o
r
e
,
“
A
c
t
i
v
i
t
y
r
e
c
o
g
n
i
t
i
o
n
u
si
n
g
c
e
l
l
p
h
o
n
e
a
c
c
e
l
e
r
o
me
t
e
r
s
,
”
AC
M
S
I
G
K
D
D
Ex
p
l
o
ra
t
i
o
n
s N
e
w
s
l
e
t
t
e
r
,
v
o
l
.
1
2
,
n
o
.
2
,
p
p
.
7
4
–
82
,
2
0
1
1
,
d
o
i
:
1
0
.
1
1
4
5
/
1
9
6
4
8
9
7
.
1
9
6
4
9
1
8
.
[
1
5
]
J.
H
o
l
l
a
n
,
E.
H
u
t
c
h
i
n
s,
a
n
d
D
.
K
i
r
sh
,
“
D
i
st
r
i
b
u
t
e
d
c
o
g
n
i
t
i
o
n
:
t
o
w
a
r
d
a
n
e
w
f
o
u
n
d
a
t
i
o
n
f
o
r
h
u
m
a
n
-
c
o
m
p
u
t
e
r
i
n
t
e
r
a
c
t
i
o
n
r
e
s
e
a
r
c
h,
”
AC
M
T
ra
n
s
a
c
t
i
o
n
s
o
n
C
o
m
p
u
t
e
r
-
H
u
m
a
n
I
n
t
e
r
a
c
t
i
o
n
,
v
o
l
.
7
,
n
o
.
2
,
p
p
.
1
7
4
–
1
9
6
,
2
0
0
0
,
d
o
i
:
1
0
.
1
1
4
5
/
3
5
3
4
8
5
.
3
5
3
4
8
7
.
[
1
6
]
L.
X
u
,
W
.
Y
a
n
g
,
Y
.
C
a
o
,
a
n
d
Q
.
Li
,
“
H
u
ma
n
a
c
t
i
v
i
t
y
r
e
c
o
g
n
i
t
i
o
n
b
a
sed
o
n
r
a
n
d
o
m
f
o
r
e
s
t
s
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
1
3
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
N
a
t
u
r
a
l
C
o
m
p
u
t
a
t
i
o
n
,
Fu
zz
y
S
y
st
e
m
s
a
n
d
K
n
o
w
l
e
d
g
e
D
i
s
c
o
v
e
ry
,
p
p
.
5
4
8
–
5
5
3
,
2
0
1
7
,
d
o
i
:
1
0
.
1
1
0
9
/
F
S
K
D
.
2
0
1
7
.
8
3
9
3
3
2
9
.
[
1
7
]
Y
.
Li
u
,
S
.
L
i
u
,
Y
.
W
a
n
g
,
F
.
L
o
m
b
a
r
d
i
,
a
n
d
J.
H
a
n
,
“
A
st
o
c
h
a
st
i
c
c
o
mp
u
t
a
t
i
o
n
a
l
mu
l
t
i
-
l
a
y
e
r
p
e
r
c
e
p
t
r
o
n
w
i
t
h
b
a
c
k
w
a
r
d
p
r
o
p
a
g
a
t
i
o
n
,
”
I
EEE
T
ra
n
s
a
c
t
i
o
n
s
o
n
C
o
m
p
u
t
e
rs
,
v
o
l
.
6
7
,
n
o
.
9
,
p
p
.
1
2
7
3
–
1
2
8
6
,
2
0
1
8
,
d
o
i
:
1
0
.
1
1
0
9
/
TC
.
2
0
1
8
.
2
8
1
7
2
3
7
.
[
1
8
]
S
.
K
i
r
a
n
y
a
z
,
T
.
I
n
c
e
,
O
.
A
b
d
e
l
j
a
b
e
r
,
O
.
A
v
c
i
,
a
n
d
M
.
G
a
b
b
o
u
j
,
“
1
D
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
f
o
r
si
g
n
a
l
p
r
o
c
e
ssi
n
g
a
p
p
l
i
c
a
t
i
o
n
s,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
I
EEE
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
Pro
c
e
ss
i
n
g
,
p
p
.
8
3
6
0
–
8
3
6
4
,
2
0
1
9
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
A
S
S
P
.
2
0
1
9
.
8
6
8
2
1
9
4
.
[
1
9
]
O
.
D
.
L
a
r
a
a
n
d
M
.
A
.
La
b
r
a
d
o
r
,
“
A
su
r
v
e
y
o
n
h
u
ma
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
w
e
a
r
a
b
l
e
s
e
n
so
r
s
,
”
I
E
EE
C
o
m
m
u
n
i
c
a
t
i
o
n
s
S
u
r
v
e
y
s
a
n
d
T
u
t
o
r
i
a
l
s
,
v
o
l
.
1
5
,
n
o
.
3
,
p
p
.
1
1
9
2
–
1
2
0
9
,
2
0
1
2
,
d
o
i
:
1
0
.
1
1
0
9
/
S
U
R
V
.
2
0
1
2
.
1
1
0
1
1
2
.
0
0
1
9
2
.
[
2
0
]
S
.
Z
h
a
n
g
,
Z
.
W
e
i
,
J.
N
i
e
,
L
.
H
u
a
n
g
,
S
.
W
a
n
g
,
a
n
d
Z.
Li
,
“
A
r
e
v
i
e
w
o
n
h
u
ma
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
v
i
si
o
n
‐
b
a
se
d
m
e
t
h
o
d
,
”
J
o
u
rn
a
l
o
f
H
e
a
l
t
h
c
a
re
E
n
g
i
n
e
e
r
i
n
g
,
v
o
l
.
2
0
1
7
,
n
o
.
1
,
p
.
3
0
9
0
3
4
3
,
2
0
1
7
,
d
o
i
:
1
0
.
1
1
5
5
/
2
0
1
7
/
3
0
9
0
3
4
3
.
[
2
1
]
A
.
R
.
M
u
r
a
l
i
d
h
a
r
a
n
,
G
.
R
i
t
h
v
i
k
,
S
.
P
r
e
m,
A
.
A
.
R
e
g
h
u
n
a
a
t
h
,
a
n
d
M
.
P
.
G
o
p
i
n
a
t
h
,
“
1
D
c
o
n
v
o
l
u
t
i
o
n
a
p
p
r
o
a
c
h
t
o
h
u
ma
n
a
c
t
i
v
i
t
y
r
e
c
o
g
n
i
t
i
o
n
u
si
n
g
se
n
s
o
r
d
a
t
a
a
n
d
c
o
mp
a
r
i
so
n
w
i
t
h
mac
h
i
n
e
l
e
a
r
n
i
n
g
a
l
g
o
r
i
t
h
ms
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
C
o
g
n
i
t
i
v
e
C
o
m
p
u
t
i
n
g
i
n
En
g
i
n
e
e
r
i
n
g
,
v
o
l
.
1
,
p
p
.
1
3
0
–
1
4
3
,
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
j
c
c
e
.
2
0
2
1
.
0
9
.
0
0
1
.
[
2
2
]
S
.
K
i
r
a
n
y
a
z
,
M
.
Z
a
b
i
h
i
,
A
.
B
.
R
a
d
,
T
.
I
n
c
e
,
R
.
H
a
mi
l
a
,
a
n
d
M
.
G
a
b
b
o
u
j
,
“
R
e
a
l
-
t
i
m
e
p
h
o
n
o
c
a
r
d
i
o
g
r
a
m
a
n
o
ma
l
y
d
e
t
e
c
t
i
o
n
b
y
a
d
a
p
t
i
v
e
1
D
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
,
”
N
e
u
ro
c
o
m
p
u
t
i
n
g
,
v
o
l
.
4
1
1
,
p
p
.
2
9
1
–
3
0
1
,
2
0
2
0
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
n
e
u
c
o
m.2
0
2
0
.
0
5
.
0
6
3
.
[
2
3
]
K
i
r
a
n
y
a
z
,
O
.
A
v
c
i
,
O
.
A
b
d
e
l
j
a
b
e
r
,
T
.
I
n
c
e
,
M
.
G
a
b
b
o
u
j
,
a
n
d
D
.
J.
I
n
ma
n
,
“
1
D
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
a
p
p
l
i
c
a
t
i
o
n
s:
a
su
r
v
e
y
,
”
Me
c
h
a
n
i
c
a
l
S
y
st
e
m
s
a
n
d
S
i
g
n
a
l
Pr
o
c
e
ssi
n
g
,
v
o
l
.
1
5
1
,
p
.
1
0
7
3
9
8
,
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
y
mss
p
.
2
0
2
0
.
1
0
7
3
9
8
.
[
2
4
]
A
.
E.
M
i
n
a
r
n
o
,
W
.
A
.
K
u
s
u
m
a
,
a
n
d
H
.
W
i
b
o
w
o
,
“
P
e
r
f
o
r
m
a
n
c
e
c
o
mp
a
r
i
so
n
a
c
t
i
v
i
t
y
r
e
c
o
g
n
i
t
i
o
n
u
si
n
g
l
o
g
i
s
t
i
c
r
e
g
r
e
ssi
o
n
a
n
d
su
p
p
o
r
t
v
e
c
t
o
r
ma
c
h
i
n
e
,
”
i
n
P
ro
c
e
e
d
i
n
g
s
o
f
t
h
e
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
I
n
t
e
l
l
i
g
e
n
t
A
u
t
o
n
o
m
o
u
s
S
y
s
t
e
m
s
,
p
p
.
1
9
–
2
4
,
2
0
2
0
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
o
I
A
S
4
9
3
1
2
.
2
0
2
0
.
9
0
8
1
8
5
8
.
[
2
5
]
Y
.
D
.
Z
h
a
n
g
,
S
.
C
.
S
a
t
a
p
a
t
h
y
,
D
.
S
.
G
u
t
t
e
r
y
,
J
.
M
.
G
o
r
r
i
z
,
a
n
d
S
.
H
.
W
a
n
g
,
“
I
mp
r
o
v
e
d
b
r
e
a
s
t
c
a
n
c
e
r
c
l
a
s
si
f
i
c
a
t
i
o
n
t
h
r
o
u
g
h
c
o
m
b
i
n
i
n
g
g
r
a
p
h
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
t
w
o
r
k
a
n
d
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
,
”
I
n
f
o
rm
a
t
i
o
n
P
ro
c
e
ss
i
n
g
and
Ma
n
a
g
e
m
e
n
t
,
v
o
l
.
5
8
,
n
o
.
2
,
p
.
1
0
2
4
3
9
,
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
p
m.
2
0
2
0
.
1
0
2
4
3
9
.
Evaluation Warning : The document was created with Spire.PDF for Python.