I
AE
S
I
n
t
e
r
n
at
ion
al
Jou
r
n
al
of
Ar
t
if
icial
I
n
t
e
ll
ig
e
n
c
e
(
I
J
-
AI
)
Vol.
14
,
No.
4
,
Augus
t
2025
,
pp.
3228
~
3240
I
S
S
N:
2252
-
8938
,
DO
I
:
10
.
11591/i
jai
.
v
14
.i
4
.
pp
32
28
-
3240
3228
Jou
r
n
al
h
omepage
:
ht
tp:
//
ij
ai
.
iaes
c
or
e
.
c
om
M
yoe
le
c
t
r
ic
gr
i
p
f
or
c
e
p
r
e
d
i
c
t
io
n
u
si
n
g
d
e
e
p
l
e
ar
n
in
g f
o
r
h
an
d
r
ob
ot
Kh
airul
Anam
1
,
4
,
Dheny
Dw
i
Ar
d
h
ian
s
yah
1
,
M
u
c
h
am
ad
Ar
if
Han
a
S
as
on
o
1
,
Ar
izal
M
u
j
ib
t
a
m
al
a
Nand
a
I
m
r
on
1
,
Na
u
f
al
Ai
n
u
r
Riz
al
1
,
M
oc
h
am
ad
E
d
owar
d
Ram
ad
h
a
n
3
,
Ar
is
Z
ain
u
l
M
u
t
t
aq
in
4
,
Clau
d
io
Cas
t
e
ll
in
i
2
,
S
u
m
ar
d
i
1
1
D
e
pa
r
tm
e
nt
of
E
le
c
tr
ic
a
l
E
ngi
ne
e
r
in
g, F
a
c
ul
ty
of
E
ngi
ne
e
r
in
g,
U
ni
ve
r
s
it
y of
J
e
mbe
r
, J
e
mbe
r
, I
ndone
s
ia
2
A
s
s
i
s
ti
ve
I
nt
e
ll
ig
e
nt
R
obot
ic
s
L
a
b, D
e
p
a
r
tm
e
nt
of
A
r
ti
f
ic
ia
l
I
nt
e
ll
ig
e
nc
e
i
n B
io
me
di
c
a
l
E
ngi
ne
e
r
in
g, F
a
c
ul
ty
of
E
ngi
ne
e
r
in
g,
F
r
ie
dr
ic
h
-
A
le
xa
nde
r
-
U
ni
ve
r
s
it
ä
t
E
r
la
nge
n
-
N
ür
nbe
r
g, E
r
la
nge
n,
G
e
r
ma
ny
3
D
e
pa
r
tm
e
nt
of
M
e
c
h
a
ni
c
a
l
E
ngi
ne
e
r
in
g, U
ni
ve
r
s
it
y of
J
e
mbe
r
,
J
e
mbe
r
, I
ndone
s
ia
4
A
r
ti
f
ic
ia
l
I
nt
e
ll
ig
e
nc
e
f
or
I
ndus
tr
ia
l
A
gr
ic
ul
tu
r
e
R
e
s
e
a
r
c
h G
r
ou
p, U
ni
ve
r
s
it
y of
J
e
mbe
r
, J
e
m
be
r
, I
ndone
s
ia
Ar
t
icle
I
n
f
o
AB
S
T
RA
CT
A
r
ti
c
le
h
is
tor
y
:
R
e
c
e
ived
F
e
b
15,
2024
R
e
vis
e
d
Apr
21,
2025
Ac
c
e
pted
J
un
8,
2025
A
rt
i
fi
c
i
al
i
n
t
el
l
i
g
en
ce
(
A
I)
h
a
s
b
een
w
i
d
el
y
ap
p
l
i
ed
i
n
t
h
e
me
d
i
ca
l
w
o
rl
d
.
O
n
e
s
u
c
h
a
p
p
l
i
ca
t
i
o
n
i
s
a
h
a
n
d
-
d
ri
v
en
ro
b
o
t
b
a
s
e
d
o
n
u
s
er
i
n
t
en
t
i
o
n
p
red
i
ct
i
o
n
.
T
h
e
p
u
r
p
o
s
e
o
f
t
h
i
s
res
earch
i
s
t
o
co
n
t
r
o
l
t
h
e
g
ri
p
s
t
ren
g
t
h
o
f
a
ro
b
o
t
b
a
s
ed
o
n
t
h
e
u
s
er’
s
i
n
t
en
t
i
o
n
b
y
p
red
i
ct
i
n
g
t
h
e
g
ri
p
s
t
re
n
g
t
h
o
f
t
h
e
u
s
er
u
s
i
n
g
d
eep
l
earn
i
n
g
an
d
el
ec
t
ro
m
y
o
g
rap
h
i
c
s
i
g
n
al
s
.
T
h
e
g
ri
p
s
t
re
n
g
t
h
o
f
t
h
e
t
arg
e
t
h
an
d
i
s
o
b
t
ai
n
e
d
fro
m
a
h
an
d
g
r
i
p
d
y
n
am
o
met
er
p
a
i
red
w
i
t
h
el
ect
r
o
my
o
g
ra
p
h
i
c
s
i
g
n
a
l
s
a
s
t
ra
i
n
i
n
g
d
at
a.
W
e
e
v
al
u
at
ed
a
co
n
v
o
l
u
t
i
o
n
al
n
eu
ra
l
n
e
t
w
o
rk
(C
N
N
)
w
i
t
h
t
w
o
d
i
ffere
n
t
arc
h
i
t
ect
u
res
.
T
h
e
i
n
p
u
t
t
o
CN
N
w
as
t
h
e
r
o
o
t
mea
n
s
q
u
are
(RMS)
an
d
mean
a
b
s
o
l
u
t
e
v
a
l
u
e
(MA
V
).
T
h
e
g
r
i
p
s
t
re
n
g
t
h
o
f
t
h
e
h
an
d
d
y
n
amo
me
t
er
w
as
u
s
ed
as
a
refere
n
ce
v
al
u
e
fo
r
a
l
o
w
-
l
ev
e
l
co
n
t
r
o
l
l
er
fo
r
t
h
e
ro
b
o
t
i
c
h
an
d
.
T
h
e
ex
p
eri
me
n
t
al
res
u
l
t
s
s
h
o
w
t
h
a
t
CN
N
s
u
cceed
ed
i
n
p
re
d
i
c
t
i
n
g
h
an
d
g
r
i
p
s
t
re
n
g
t
h
a
n
d
co
n
t
r
o
l
l
i
n
g
g
ri
p
s
t
re
n
g
t
h
w
i
t
h
a
r
o
o
t
mea
n
s
q
u
are
e
rro
r
(
RMSE
)
o
f
2
.
3
5
N
u
s
i
n
g
t
h
e
RMS
feat
u
re.
A
co
mp
a
ri
s
o
n
w
i
t
h
a
s
t
a
t
e
-
of
-
the
-
ar
t
reg
res
s
i
o
n
met
h
o
d
al
s
o
s
h
o
w
s
t
h
a
t
a
CN
N
can
b
et
t
er
p
red
i
ct
t
h
e
g
r
i
p
s
t
re
n
g
t
h
.
K
e
y
w
o
r
d
s
:
As
s
is
ti
ve
r
obot
De
e
p
lea
r
ning
Gr
ip
f
o
r
c
e
Ha
nd
r
obot
M
yoe
lec
tr
ic
Th
i
s
i
s
a
n
o
p
en
a
c
ces
s
a
r
t
i
c
l
e
u
n
d
e
r
t
h
e
CC
B
Y
-
SA
l
i
ce
n
s
e.
C
or
r
e
s
pon
din
g
A
u
th
or
:
Kha
ir
ul
Ana
m
De
pa
r
tm
e
nt
of
E
lec
tr
ica
l
E
nginee
r
ing,
F
a
c
ult
y
o
f
E
nginee
r
ing
,
Unive
r
s
it
y
of
J
e
mber
Ka
li
manta
n
S
tr
e
e
t,
T
e
ga
lbot
o
No
.
37
,
K
r
a
jan
T
im
u
r
,
S
umber
s
a
r
i
,
J
e
mber
,
I
ndone
s
ia
E
mail:
kha
ir
u
l@unej.
a
c
.
id
1.
I
NT
RODU
C
T
I
ON
T
he
c
ombi
na
ti
on
of
a
r
ti
f
icia
l
int
e
ll
igenc
e
(
A
I
)
a
n
d
bios
e
ns
or
s
withi
n
a
s
s
is
ti
ve
s
ys
tems
ha
s
yielde
d
pr
omi
s
ing
outcome
s
a
c
r
os
s
va
r
ious
r
e
s
e
a
r
c
h
e
nde
a
vor
s
.
A
I
-
e
nha
nc
e
d
bios
e
ns
or
s
ha
ve
de
mons
tr
a
ted
potential
f
or
r
a
pid
diagnos
ti
c
s
,
pr
e
c
is
ion
ther
a
pe
uti
c
s
,
a
n
d
dis
e
a
s
e
mana
ge
ment
[
1]
.
T
he
s
e
tec
hnologi
e
s
leve
r
a
ge
mac
hine
lea
r
ning,
ne
ur
a
l
ne
twor
ks
,
a
nd
other
AI
t
e
c
hniques
to
im
pr
ove
bios
e
ns
or
f
unc
ti
ona
li
ty,
c
on
ne
c
ti
vit
y,
a
nd
point
-
of
-
c
a
r
e
a
dopti
on
[
2]
.
W
e
a
r
a
ble
b
ios
e
ns
ing
tec
hnologi
e
s
,
e
mpowe
r
e
d
by
AI
,
e
na
ble
the
m
onit
or
ing
of
phys
iol
ogica
l
s
ignals
a
nd
a
id
in
dis
e
a
s
e
diagno
s
is
,
s
uppor
ti
ng
the
tr
e
nd
towa
r
d
pe
r
s
ona
li
z
e
d
m
e
dicine
[
3]
.
T
he
c
ombi
na
ti
on
of
AI
with
s
e
ns
ing
tec
hnology
ha
s
led
to
the
de
ve
lopm
e
nt
of
int
e
ll
igent
bios
e
ns
or
s
c
a
pa
ble
of
r
a
pid
tar
ge
t
de
tec
ti
on
with
high
s
e
ns
it
ivi
ty,
a
c
c
ur
a
c
y,
a
nd
pr
e
c
is
ion
[
4]
.
J
in
e
t
al
.
[
5]
de
lve
int
o
the
c
ha
ll
e
nge
s
a
nd
pr
os
pe
c
ts
a
s
s
o
c
iate
d
with
AI
-
dr
i
ve
n
bios
e
ns
or
s
,
highl
ight
ing
the
s
igni
f
ica
nc
e
of
mate
r
ial
a
dva
nc
e
ments
,
bior
e
c
ognit
ion
c
omponents
,
a
nd
da
ta
pr
oc
e
s
s
ing
tec
hnique
s
.
T
he
s
e
ins
ight
s
of
f
e
r
va
luabl
e
guidanc
e
f
or
the
f
utur
e
e
volut
ion
o
f
AI
-
ba
s
e
d
bios
e
ns
or
s
tailor
e
d
f
or
a
s
s
is
ti
ve
a
ppli
c
a
ti
ons
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
Ar
ti
f
I
ntell
I
S
S
N:
2252
-
8938
M
y
oe
lec
tr
ic
gr
ip
for
c
e
pr
e
diction
us
ing
de
e
p
lear
n
ing
for
hand
r
obot
(
K
hair
ul
A
nam
)
3229
E
lec
tr
omyogr
a
phy
(
E
M
G)
is
a
method
f
o
r
r
e
c
or
di
ng
mus
c
le
s
ignals
.
E
M
G
ha
s
va
r
ious
a
ppli
c
a
ti
ons
,
including
c
ontr
oll
ing
p
r
os
thetic
r
obots
f
or
a
mput
e
e
pa
ti
e
nts
to
i
mpr
ove
the
r
obot
-
us
e
r
int
e
r
a
c
ti
on
[
6]
–
[
11]
.
Ana
m
a
nd
Al
-
J
umaily
[
6]
f
oc
us
e
d
on
de
ve
lopi
ng
a
n
a
mput
a
ti
on
r
obot
that
moves
a
c
c
or
ding
to
th
e
us
e
r
’
s
wis
he
s
,
pa
ying
a
tt
e
nti
on
to
the
us
e
r
’
s
c
omf
or
t
a
nd
a
s
if
the
r
obot
we
r
e
a
pa
r
t
of
his
body
.
A
pr
os
thetic
r
obot
is
idea
ll
y
us
e
d
to
r
e
plac
e
the
pa
ti
e
nt's
ha
nd
a
nd
move
s
moot
hly
a
nd
with
s
pe
c
if
ic
s
tr
e
ngth
a
c
c
or
di
ng
to
the
us
e
r
’
s
int
e
nti
on
[
12]
.
T
o
e
ns
ur
e
that
a
r
obot's
ha
nd
c
a
n
hold
objec
ts
c
or
r
e
c
tl
y
a
nd
pr
e
c
is
e
ly
a
c
c
or
din
g
to
gr
ip
s
tr
e
ngth,
it
is
c
r
uc
ial
to
p
r
e
dict
the
g
r
ip
s
tr
e
ngth.
T
he
r
e
ha
ve
be
e
n
va
r
ious
a
ppr
oa
c
he
s
pr
opos
e
d
to
pr
e
dict
gr
ip
s
tr
e
ngth.
L
o
e
t
al.
[
13]
inves
ti
ga
ted
the
us
e
of
gr
ip
s
tr
e
ngth
to
pr
e
dict
other
ha
nd
e
xe
r
ti
ons
,
f
indi
n
g
it
les
s
e
f
f
e
c
ti
ve
f
o
r
pa
lm
a
r
pinch
a
nd
thum
b
pr
e
s
s
.
L
v
e
t
al.
[
14
]
pr
opos
e
d
a
method
ba
s
e
d
on
s
ur
f
a
c
e
E
M
G
s
ignals
,
opti
m
izin
g
a
s
uppor
t
ve
c
tor
r
e
gr
e
s
s
ion
model
thr
ough
the
s
pa
r
r
ow
s
e
a
r
c
h
a
lgor
it
h
m
to
a
c
c
ur
a
tely
p
r
e
dict
ha
nd
gr
ip
s
tr
e
ngth.
S
a
ya
diza
de
h
e
t
al.
[
15
]
uti
li
z
e
d
a
r
ti
f
icia
l
ne
u
r
a
l
n
e
twor
ks
to
pr
e
dict
gr
ip
a
nd
pinch
s
tr
e
ngth
ba
s
e
d
on
ha
nd
a
nthr
opometr
ic
pa
r
a
mete
r
s
,
identi
f
ying
ke
y
pr
e
dictor
s
s
uc
h
a
s
ha
nd
length
,
width
,
a
nd
s
ha
pe
index.
C
hihi
et
al.
[
16
]
us
e
d
a
tec
hnique
ba
s
e
d
on
the
no
nli
ne
a
r
Ha
mm
e
r
s
tein
-
W
iene
r
model.
S
ome
r
e
s
e
a
r
c
he
r
s
a
r
e
a
ls
o
s
tar
ti
ng
to
us
e
de
e
p
lea
r
ning
to
pr
e
dict
gr
ip
s
tr
e
ng
th.
S
u
e
t
al
.
[
17]
pr
opos
e
d
a
c
onvolut
ion
a
l
ne
ur
a
l
ne
twor
k
(
C
NN
)
to
p
r
e
dict
the
s
tr
e
ngth
of
the
E
M
G
s
ignal.
How
e
ve
r
,
thi
s
s
tudy
did
no
t
s
pe
c
if
ica
ll
y
f
oc
us
on
ha
nd
-
gr
ip
s
tr
e
ngth.
Hw
a
ng
e
t
al.
[
18
]
pr
e
s
e
nted
de
e
p
ne
ur
a
l
ne
twor
ks
that
c
a
n
pr
e
dict
ha
nd
gr
ip
s
tr
e
ngth,
but
the
r
e
s
ult
s
ha
ve
not
ye
t
be
e
n
us
e
d
to
c
ontr
ol
r
obots
dir
e
c
tl
y
.
I
t
is
ne
c
e
s
s
a
r
y
to
f
ur
ther
e
va
luate
the
im
pleme
ntation
o
f
f
or
c
e
pr
e
diction
on
r
obo
ti
c
ha
nds
.
S
e
ve
r
a
l
inves
ti
ga
ti
ons
ha
ve
de
lved
int
o
uti
li
z
ing
de
e
p
lea
r
ning
models
to
f
or
e
c
a
s
t
gr
ip
s
tr
e
ngth
thr
ough
E
M
G
s
ignals
.
Xu
e
t
al.
[
19]
i
nt
r
oduc
e
d
im
pe
da
nc
e
s
ignals
to
p
r
e
dict
g
r
ip
f
or
c
e
,
a
c
hieving
high
a
c
c
ur
a
c
y
wi
th
a
long
s
hor
t
-
ter
m
memor
y
(
L
S
T
M
)
model
.
J
iang
e
t
al.
[
20]
de
vis
e
d
a
n
a
da
pti
ve
ne
ur
a
l
f
uz
z
y
inf
e
r
e
nc
e
s
ys
tem
e
mpl
oying
s
ur
f
a
c
e
E
M
G
s
ignals
,
e
f
f
e
c
ti
ve
ly
p
r
e
di
c
ti
ng
gr
ip
s
tr
e
n
gth
a
nd
of
f
e
r
ing
ins
ight
s
int
o
r
e
ha
bil
it
a
ti
ve
ther
a
py.
M
a
e
t
al.
[
21
]
uti
li
z
e
d
a
ge
ne
e
xpr
e
s
s
ion
pr
ogr
a
mm
ing
a
lgor
it
hm
a
nd
a
ba
c
k
pr
opa
ga
ti
o
n
ne
ur
a
l
ne
twor
k
to
c
ons
tr
uc
t
a
p
r
e
diction
model
f
o
r
g
r
a
s
ping
f
or
c
e
ba
s
e
d
on
s
E
M
G
s
ignals
,
a
c
hieving
im
pr
e
s
s
ive
a
c
c
ur
a
c
y.
How
e
ve
r
,
the
r
e
s
ult
s
we
r
e
no
t
di
r
e
c
tl
y
us
e
d
to
c
ont
r
ol
the
r
obot
.
T
he
s
e
s
tudi
e
s
de
mons
tr
a
te
the
potential
of
de
e
p
lea
r
ning
models
in
a
c
c
ur
a
tely
f
o
r
e
c
a
s
ti
ng
gr
ip
s
tr
e
ngth
thr
ough
E
M
G
s
ignals
.
Ho
we
ve
r
,
it
s
hould
be
noted
that
the
r
e
s
ult
s
we
r
e
not
dir
e
c
tl
y
a
ppli
e
d
to
r
obo
t
c
ontr
ol
.
T
his
a
r
ti
c
le
a
i
ms
to
de
s
ign
a
gr
ip
s
tr
e
ngth
c
ontr
ol
s
ys
tem
f
or
ha
nd
r
obots
us
ing
de
e
p
lea
r
ning,
s
pe
c
if
ica
ll
y
the
C
NN
.
T
his
r
e
s
e
a
r
c
h
pr
e
s
e
nts
a
no
ve
l
f
r
a
mew
or
k
f
or
r
oboti
c
ha
nd
c
ont
r
ol
thr
ough
g
r
a
s
p
f
or
c
e
pr
e
diction,
a
dva
nc
ing
the
s
tate
-
of
-
th
e
-
a
r
t
in
d
e
xter
ous
manipulation.
T
he
f
r
a
mew
or
k
inco
r
p
or
a
tes
a
c
ompr
e
he
ns
ive
c
ompar
a
ti
ve
a
na
lys
is
of
de
e
p
lea
r
ning
a
r
c
hit
e
c
tur
e
s
,
s
pe
c
if
ica
ll
y
e
va
luating
va
r
io
us
C
NN
models
a
ga
ins
t
tr
a
dit
ional
a
ppr
oa
c
he
s
,
including
L
S
T
M
ne
twor
ks
a
nd
c
las
s
i
c
a
l
mac
hine
lea
r
ni
ng
a
lgor
it
hms
s
uc
h
a
s
r
a
ndom
f
or
e
s
t
(
R
F
)
,
k
-
ne
a
r
e
s
t
n
e
ighbor
s
(
k
-
NN
)
,
a
nd
de
c
is
ion
tr
e
e
s
(
DT
)
.
T
he
e
xpe
r
im
e
nta
l
r
e
s
ult
s
de
mons
tr
a
te
the
f
r
a
mew
or
k's
e
f
f
e
c
ti
ve
ne
s
s
in
e
n
ha
nc
ing
gr
ip
pr
e
c
is
ion
a
nd
c
ont
r
ol
e
f
f
icie
nc
y,
p
r
ovidi
ng
va
luable
ins
ight
s
f
or
the
de
ve
lopm
e
nt
of
mo
r
e
s
ophis
ti
c
a
ted
r
oboti
c
manipulation
s
ys
tems
.
I
n
a
ddit
ion
to
it
s
a
ppli
c
a
ti
on
in
r
obot
ic
c
ontr
ol
,
C
NN
s
ha
ve
be
e
n
wide
ly
uti
li
z
e
d
in
other
do
mains
.
E
xa
mpl
e
s
include
e
a
r
ly
s
tr
oke
dis
e
a
s
e
pr
e
diction
[
22]
a
nd
de
tec
ti
ng
s
tu
de
nt
a
tt
e
nti
on
leve
ls
[
23]
.
T
he
s
e
e
xa
mpl
e
s
high
li
ght
the
ve
r
s
a
ti
li
ty
of
C
NN
s
in
a
dd
r
e
s
s
ing
diver
s
e
c
ha
ll
e
nge
s
a
c
r
os
s
f
ields
.
T
he
a
r
ti
c
le's
s
tr
uc
tur
e
is
a
s
f
oll
ows
.
S
e
c
ti
on
2
dis
c
u
s
s
e
s
the
methods
us
e
d
in
thi
s
s
tudy.
S
e
c
ti
on
3
pr
e
s
e
nts
the
r
e
s
ult
s
a
nd
dis
c
us
s
ion.
F
inally,
the
a
r
ti
c
le
c
onc
ludes
with
ke
y
f
indi
ngs
a
nd
f
utur
e
dir
e
c
ti
o
ns
.
2.
M
E
T
HO
DS
T
his
r
e
s
e
a
r
c
h
a
im
s
to
de
s
ign
a
r
obot
c
ontr
ol
s
ys
tem
ba
s
e
d
on
pr
e
dictions
o
f
the
us
e
r
's
gr
ip
s
tr
e
ngth
ba
s
e
d
on
mus
c
le
s
ignals
us
ing
de
e
p
lea
r
ning
meth
ods
.
T
he
de
e
p
lea
r
ning
method
us
e
d
is
a
C
NN
.
T
h
e
ge
ne
r
a
l
de
s
ign
of
the
s
ys
tem
is
s
hown
in
F
igur
e
1.
T
he
pr
opos
e
d
c
ontr
ol
s
ys
tem
a
ppe
a
r
s
to
be
a
n
ope
n
loop
be
c
a
us
e
ther
e
a
r
e
no
pr
e
s
s
ur
e
s
e
ns
or
s
on
the
r
obo
t
o
r
the
r
obot’
s
f
inger
ti
ps
.
How
e
ve
r
,
the
de
tec
ted
s
tr
e
ngt
h
s
e
ns
or
c
omes
dir
e
c
tl
y
f
r
om
the
us
e
r
,
na
mely
f
r
om
the
mu
s
c
le
s
ignals
de
c
ode
d
by
C
NN
.
De
tails
o
f
e
a
c
h
s
tage
will
be
e
xplaine
d
in
the
f
oll
owing
s
e
c
ti
ons
.
2.
1.
Ac
q
u
is
it
ion
an
d
p
r
e
-
p
r
oc
e
s
s
in
g
d
at
a
T
he
M
yo
a
r
mband
c
oll
e
c
ted
mus
c
le
s
ignals
,
or
E
M
G.
T
he
M
yo
a
r
mband
wa
s
a
tt
a
c
he
d
to
the
pa
r
ti
c
ipant's
f
or
e
a
r
m
while
the
ha
nd
g
r
ipped
t
he
ha
nd
dyna
mom
e
ter
.
Da
ta
we
r
e
c
oll
e
c
ted
f
r
om
f
ive
r
e
s
ponde
nts
who
we
r
e
ins
tr
uc
ted
to
move
th
e
ir
ha
nds
thr
oug
h
ope
ning
a
nd
g
r
a
s
ping
mot
ions
a
t
pr
e
de
ter
mi
ne
d
int
e
r
va
ls
of
15
s
e
c
onds
f
or
e
a
c
h
s
tr
e
ngth
pa
r
a
mete
r
.
T
he
s
tudy
pa
r
ti
c
ipants
we
r
e
m
e
n
a
ge
d
20
to
25
ye
a
r
s
with
good
ha
nd
mus
c
le
c
ondit
ion
b
a
s
e
d
on
the
diame
ter
of
their
a
r
ms
.
T
his
r
e
s
e
a
r
c
h
include
d
r
e
s
ponde
nt
s
who
ha
d
phys
ica
l
a
nd
menta
l
he
a
lt
h
without
a
his
tor
y
of
i
ll
ne
s
s
.
T
he
E
M
G
de
vice
wa
s
ins
talled
on
the
f
or
e
a
r
m
of
the
r
e
s
ponde
nt's
r
ight
ha
nd,
whic
h
he
ld
the
ha
nd
dyna
mom
e
ter
.
E
f
f
o
r
ts
we
r
e
made
to
e
ns
ur
e
that
the
r
e
s
ponde
nts
w
e
r
e
a
s
c
omf
or
table
a
s
pos
s
ibl
e
s
o
they
c
ould
f
oc
us
on
movi
ng
their
ha
nds
.
T
he
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
Ar
ti
f
I
ntell
,
Vol.
14
,
No.
4
,
Augus
t
2025
:
322
8
-
3240
3230
c
ondit
ions
of
the
r
e
s
ponde
nts
dur
ing
da
ta
c
oll
e
c
ti
on
a
r
e
s
hown
in
F
igur
e
2.
T
his
da
ta
c
oll
e
c
ti
on
pr
oc
e
dur
e
wa
s
a
ppr
ove
d
by
the
E
thi
c
s
C
omm
is
s
ion
of
the
Unive
r
s
it
y
o
f
J
e
mber
unde
r
number
960/UN25.
8/KE
P
K/DL
/202
0.
T
he
s
teps
in
p
r
e
pr
oc
e
s
s
ing
E
M
G
s
ignal
da
ta
a
r
e
s
ignal
f
i
lt
e
r
ing
a
nd
windowing
,
that
is
,
taking
a
s
ignal
a
t
a
c
e
r
tain
ti
me.
T
he
E
M
G
s
ignal
f
il
ter
ing
pr
oc
e
s
s
us
e
s
a
ba
ndpa
s
s
f
il
te
r
by
e
nter
ing
the
up
pe
r
li
mi
t
a
nd
lowe
r
li
mi
t
va
lue
pa
r
a
mete
r
s
a
nd
a
notch
f
il
ter
to
ove
r
c
ome
the
dis
tur
ba
nc
e
of
the
mains
volt
a
ge
.
A
ba
ndpa
s
s
f
il
ter
e
ns
ur
e
s
that
the
s
ignal
be
ing
pr
oc
e
s
s
e
d
is
a
n
E
M
G
s
ign
a
l,
whic
h
us
ua
ll
y
r
a
nge
s
f
r
om
10
to
500
Hz
.
T
he
s
tr
e
ngth
pr
e
diction
s
ys
tem
is
a
n
ove
r
view
of
the
da
ta
pr
oc
e
s
s
ing
s
ys
tem
s
hown
in
F
igur
e
3.
T
he
pr
e
diction
s
ys
tem
c
ons
is
ts
of
thr
e
e
s
tage
s
:
da
ta
a
c
quis
it
ion,
p
r
e
pr
oc
e
s
s
ing,
a
nd
pr
e
dictio
n.
Da
ta
a
c
quis
it
ion
c
oll
e
c
ts
mus
c
le
e
lec
tr
ica
l
a
c
ti
vit
y
a
nd
ha
ndgr
ip
s
tr
e
ngth
(
N)
.
P
r
e
pr
oc
e
s
s
ing
invol
ve
s
E
M
G
s
ignal
f
il
ter
ing
(
ba
ndpa
s
s
a
nd
notch
f
i
lt
e
r
s
)
,
windowing
f
or
s
tr
uc
tur
e
d
s
a
mpl
ing,
a
nd
f
e
a
tur
e
e
xt
r
a
c
ti
on
us
ing
r
oot
mea
n
s
qua
r
e
(
R
M
S
)
a
nd
mea
n
a
bs
olut
e
va
lue
(
M
AV
)
.
F
inally
,
C
NN
is
us
e
d
to
pr
e
dict
E
M
G
da
ta,
g
e
ne
r
a
ti
ng
a
n
a
c
c
ur
a
te
pr
e
diction
model
.
F
igur
e
1
.
T
he
pr
opos
e
d
c
ontr
ol
s
ys
tem
F
igur
e
2.
Da
ta
c
oll
e
c
ti
on
F
igur
e
3.
Da
ta
pr
oc
e
s
s
ing
2.
2.
F
e
a
t
u
r
e
e
xt
r
ac
t
ion
F
e
a
tur
e
e
xtr
a
c
ti
on
is
pe
r
f
or
med
be
f
or
e
the
f
o
r
e
c
a
s
ti
ng
s
tage
us
ing
a
C
NN
.
T
he
f
e
a
tur
e
e
xtr
a
c
ti
on
us
e
d
in
thi
s
inves
ti
ga
ti
on
is
the
R
M
S
a
nd
M
AV
.
Apa
r
t
f
r
om
that
,
r
a
w
s
ignals
will
a
ls
o
be
tes
ted
without
going
thr
ough
the
e
xtr
a
c
ti
on
f
e
a
tur
e
.
T
h
e
ma
t
he
mat
i
c
a
l
f
o
r
m
ul
a
f
o
r
t
he
R
M
S
f
e
a
t
u
r
e
e
xt
r
a
c
t
i
on
is
s
h
own
i
n
(
1)
.
=
√
1
∫
(
)
2
+
(
1)
T
he
M
AV
e
xtr
a
c
ti
on
f
e
a
tur
e
is
wide
ly
us
e
d
to
p
r
oc
e
s
s
E
M
G
s
ignals
digi
tally
.
M
AV
de
s
c
r
ibes
the
s
ignal
f
e
a
tur
e
s
us
ing
the
f
or
mul
a
f
o
r
a
ve
r
a
ge
a
nd
a
bs
olut
e
va
lues
.
T
he
mathe
matica
l
f
or
mul
a
f
o
r
t
he
M
AV
f
e
a
tur
e
e
xtr
a
c
ti
on
is
s
hown
in
(
2
)
.
=
1
∑
|
|
=
1
(
2)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
Ar
ti
f
I
ntell
I
S
S
N:
2252
-
8938
M
y
oe
lec
tr
ic
gr
ip
for
c
e
pr
e
diction
us
ing
de
e
p
lear
n
ing
for
hand
r
obot
(
K
hair
ul
A
nam
)
3231
2.
3.
CNN
ar
c
h
it
e
c
t
u
r
e
d
e
s
ign
f
or
gr
ip
s
t
r
e
n
gt
h
p
r
e
d
ict
ion
T
h
e
C
NN
a
r
c
h
i
tec
tu
r
e
is
c
om
p
r
is
e
d
o
f
tw
o
ma
in
p
a
r
t
s
:
f
e
a
tu
r
e
l
e
a
r
n
in
g
a
n
d
c
las
s
i
f
i
c
a
ti
on
.
T
he
f
e
a
tu
r
e
l
e
a
r
ni
ng
p
r
oc
e
s
s
in
vo
lv
e
s
a
s
e
r
i
e
s
o
f
la
ye
r
s
,
i
nc
lu
d
in
g
c
o
nv
ol
ut
io
na
l
a
n
d
p
oo
li
ng
la
ye
r
s
.
T
he
c
las
s
i
f
ic
a
t
io
n
p
r
oc
e
s
s
i
nc
lu
de
s
f
l
a
t
te
ne
d
a
n
d
f
u
ll
y
c
on
ne
c
te
d
l
a
y
e
r
s
.
T
h
is
r
e
s
e
a
r
c
h
i
nve
s
t
iga
tes
tw
o
C
NN
a
r
c
hi
te
c
t
u
r
e
s
,
e
a
c
h
w
i
th
pa
r
a
me
te
r
s
s
u
c
h
a
s
in
pu
t
s
ha
p
e
,
f
i
lt
e
r
,
ke
r
ne
l
s
i
z
e
,
p
oo
li
ng
s
iz
e
,
a
n
d
f
u
l
ly
c
on
ne
c
te
d
l
a
ye
r
.
T
h
e
s
t
ud
y
u
t
il
iz
e
s
a
o
ne
-
di
me
ns
i
on
a
l
C
NN
(
1
D
C
N
N
)
,
a
n
d
the
p
a
r
a
met
e
r
va
l
ue
s
a
r
e
p
r
e
s
e
n
ted
in
T
a
b
le
1
.
T
a
ble
1.
C
NN
a
r
c
hit
e
c
tur
e
c
onf
igu
r
a
ti
on
C
onvolut
io
n
l
a
ye
r
K
e
r
ne
l
s
iz
e
P
ool
in
g s
iz
e
A
r
c
hi
te
c
tu
r
e
L
a
ye
r
1
L
a
ye
r
2
L
a
ye
r
3
L
a
ye
r
1
L
a
ye
r
2
L
a
ye
r
3
L
a
ye
r
1
L
a
ye
r
2
L
a
ye
r
3
C
N
N
s
1
128
64
32
4
2
1
2
1
1
C
N
N
s
2
128
64
-
4
2
-
2
1
-
I
n
T
a
ble
1
,
t
he
va
lues
of
the
a
r
c
hit
e
c
tur
a
l
pa
r
a
mete
r
s
r
e
main
the
s
a
me,
a
nd
the
only
dif
f
e
r
e
nc
e
is
in
the
input
s
ha
pe
.
T
he
input
s
ha
pe
of
the
r
a
w
da
t
a
is
40
.
8,
whic
h
is
obtaine
d
f
r
om
the
windowin
g
r
e
s
ult
s
.
On
the
other
ha
nd,
the
input
s
ha
pe
o
f
the
R
M
S
a
nd
M
AV
da
ta
is
8
.
1,
whic
h
is
obtaine
d
a
f
ter
r
e
s
h
a
ping
the
windowing
da
ta.
T
his
is
ne
c
e
s
s
a
r
y
be
c
a
us
e
f
e
a
tur
e
e
xtr
a
c
ti
on
us
ing
R
M
S
a
nd
M
AV
is
not
s
e
que
nti
a
l.
T
he
or
igi
na
l
windowing
da
ta
wa
s
a
3
-
dim
e
ns
ional
matr
ix
,
a
nd
it
be
c
a
me
two
-
dim
e
ns
ional
due
to
th
e
los
s
of
s
e
que
nc
e
.
T
he
r
e
f
or
e
,
r
e
s
ha
ping
is
ne
c
e
s
s
a
r
y
to
c
h
a
nge
the
R
M
S
a
nd
M
AV
da
ta
matr
ice
s
to
matc
h
t
he
C
NN
.
T
he
C
NN
a
r
c
hit
e
c
tur
e
c
a
n
be
de
s
c
r
ibed
in
F
igu
r
e
s
4
a
nd
5.
F
igur
e
4
s
hows
8
de
pth
laye
r
s
s
tar
ti
ng
f
r
om
input
s
ha
pe
,
c
onvolut
ion
1
,
pooli
ng
1,
c
onvolut
ion
2,
p
ooli
ng
2,
c
onvolut
ion
3
,
poo
li
ng
3
,
f
latten,
a
nd
de
ns
e
,
a
nd
F
igur
e
5
s
hows
the
a
r
c
hit
e
c
tur
e
2
r
a
w
da
ta
ha
s
2
c
onvolut
ion
laye
r
s
,
2
pooli
ng
laye
r
s
,
f
latten,
a
nd
de
ns
e
.
F
igur
e
4.
C
NN
a
r
c
hit
e
c
tur
e
1
F
igur
e
5.
C
NN
a
r
c
hit
e
c
tur
e
2
2.
4.
P
r
e
d
ic
t
h
an
d
gr
ip
s
t
r
e
n
gt
h
wi
t
h
CN
N
T
he
input
f
or
C
NN
is
f
r
om
the
f
e
a
tur
e
e
xt
r
a
c
ti
on
pr
oc
e
s
s
on
the
E
M
G
s
ignal
us
ing
R
M
S
a
nd
M
AV
.
T
his
r
e
s
e
a
r
c
h
uti
li
z
e
d
a
1D
C
NN
with
s
e
ve
r
a
l
lay
e
r
s
:
the
input
,
c
onvolut
ional,
s
a
mpl
ing,
a
nd
outpu
t
laye
r
s
.
A
s
c
he
matic
r
e
pr
e
s
e
ntation
of
the
C
NN
method
f
or
p
r
e
dicting
mus
c
le
gr
ip
s
tr
e
ngth
is
s
hown
in
the
F
igu
r
e
6
il
lus
tr
a
tes
the
pr
e
diction
methodology
us
ing
r
a
w
E
M
G
da
ta
a
nd
C
NN
.
T
he
M
yo
a
r
mband
r
e
c
or
ds
mus
c
le
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
Ar
ti
f
I
ntell
,
Vol.
14
,
No.
4
,
Augus
t
2025
:
322
8
-
3240
3232
e
lec
tr
ica
l
a
c
ti
vit
y
f
r
om
e
ight
s
e
ns
or
c
ha
nne
ls
,
w
hich
unde
r
go
ba
ndpa
s
s
a
nd
notch
f
il
ter
ing
be
f
or
e
be
ing
windowe
d.
F
e
a
tur
e
s
a
r
e
e
xtr
a
c
ted
us
ing
R
M
S
a
nd
M
AV
,
s
e
r
ving
a
s
C
NN
input
.
T
he
C
NN
pr
oc
e
s
s
include
s
c
onvolut
ion,
pooli
ng,
a
nd
f
ull
y
c
onne
c
ted
laye
r
s
,
tr
a
ns
f
or
mi
ng
da
ta
f
r
o
m
two
to
thr
e
e
dim
e
ns
ions
(
x,
y,
z
)
with
f
il
ter
s
(
z
-
a
xis
)
a
nd
ke
r
ne
l
s
ize
a
djus
tm
e
nts
.
T
he
r
e
s
ult
ing
f
e
a
tur
e
map
pr
e
dicts
ha
ndgr
ip
s
tr
e
ngt
h,
whic
h
is
r
e
c
or
de
d
a
longs
ide
E
M
G
s
ignals
.
F
igur
e
6.
S
c
he
matic
r
e
pr
e
s
e
ntation
of
the
pr
e
dictio
n
of
the
gr
ip
f
or
c
e
2.
5.
L
ow
-
c
os
t
p
r
os
t
h
e
t
ic
h
an
d
A
ha
nd
pr
os
thetic
r
obo
t
is
a
r
oboti
c
ha
nd
that
is
us
e
d
f
or
g
r
ippi
ng.
T
his
type
of
pr
os
thetic
is
de
s
igned
to
e
xc
e
l
a
t
gr
a
s
ping
r
a
ther
than
manipulating
tas
ks
.
I
t
f
e
a
tur
e
s
hig
h
de
xter
it
y
,
s
ophis
ti
c
a
ted
s
e
ns
or
s
,
a
nd
a
dva
nc
e
d
c
ontr
ol
s
tr
a
tegie
s
.
F
ive
s
e
r
vo
mot
o
r
s
a
r
e
us
e
d
on
the
f
ive
r
obot
f
inger
s
with
the
A
r
duino,
a
s
s
hown
in
F
igu
r
e
7
.
I
n
th
is
f
igu
r
e
,
the
c
onf
igu
r
a
ti
on
of
the
s
e
r
vo
mot
or
pins
on
the
r
obot's
f
inger
is
a
s
f
oll
ows
:
thum
b
on
pin
3,
index
f
inger
on
pin
4,
m
iddl
e
f
ing
e
r
on
pin
5,
r
ing
f
inger
on
pin
6
,
a
nd
li
tt
le
f
inger
o
n
pin
7.
T
he
M
yo
Ar
mband
dongle
c
onne
c
ts
to
a
P
C
via
B
luetooth,
whic
h
is
s
e
r
ially
l
inked
to
a
n
Ar
duino
Uno
f
or
tr
a
ns
mi
tt
ing
s
e
r
vo
mot
or
a
ngle
c
ontr
ol
d
a
ta.
F
ive
s
e
r
vomot
or
s
s
ha
r
e
pa
r
a
ll
e
l
VC
C
a
nd
gr
ound
c
on
ne
c
ti
ons
,
powe
r
e
d
thr
ough
a
n
L
M
2596
s
tep
-
down
modul
e
with
a
3
-
c
e
ll
2200
mAh
L
iP
o
ba
tt
e
r
y.
T
he
11
.
1
V
ba
tt
e
r
y
input
is
r
e
gulate
d
to
5
V
,
matc
hing
the
s
e
r
vo
mot
or
's
ope
r
a
ti
ng
volt
a
ge
.
F
igur
e
7.
E
lec
tr
onic
c
ir
c
uit
de
s
ign
of
ha
nd
r
obot
3.
RE
S
UL
T
S
AN
D
DI
S
CU
S
S
I
ON
3.
1.
Of
f
li
n
e
t
e
s
t
in
g
r
e
s
u
lt
s
o
f
CN
N
d
a
t
a
RA
W
,
RM
S
,
an
d
M
AV
a
r
c
h
it
e
c
t
u
r
e
Onc
e
the
pa
r
a
mete
r
s
of
the
C
NN
a
nd
it
s
a
r
c
hit
e
c
tur
e
a
r
e
de
f
ined,
e
xpe
r
im
e
nts
a
r
e
c
onduc
ted
on
the
R
AW
,
R
M
S
,
a
nd
M
AV
da
ta
to
identif
y
the
m
os
t
a
c
c
ur
a
te
pr
e
dictive
r
e
s
ult
s
f
r
om
the
p
r
e
-
de
ter
mi
ne
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
Ar
ti
f
I
ntell
I
S
S
N:
2252
-
8938
M
y
oe
lec
tr
ic
gr
ip
for
c
e
pr
e
diction
us
ing
de
e
p
lear
n
ing
for
hand
r
obot
(
K
hair
ul
A
nam
)
3233
a
r
c
hit
e
c
tur
e
.
T
o
a
c
hieve
thi
s
,
a
t
r
a
ini
ng
a
nd
tes
ti
ng
r
a
ti
o
of
7:3
wa
s
uti
li
z
e
d.
T
he
outcome
s
of
the
e
xp
e
r
im
e
nts
c
onduc
ted
on
the
R
AW
,
R
M
S
,
a
nd
M
AV
da
ta
f
r
o
m
the
two
a
r
c
h
it
e
c
tur
e
s
,
na
mely
a
r
c
hit
e
c
tur
e
1
a
nd
a
r
c
hit
e
c
tur
e
2,
a
r
e
pr
e
s
e
nted
in
T
a
ble
2
.
B
a
s
e
d
on
the
da
ta
obtaine
d,
a
c
ompar
is
on
ha
s
be
e
n
made
of
thr
e
e
pr
e
-
pr
oc
e
s
s
ing
methods
:
r
a
w
da
ta,
R
M
S
,
a
nd
M
AV
.
W
it
h
r
a
w
da
ta,
the
C
NN
1
a
r
c
hit
e
c
tur
e
a
c
hieve
s
a
n
a
ve
r
a
ge
a
c
c
ur
a
c
y
that
e
xc
e
e
d
s
that
of
the
C
NN
2
a
r
c
hit
e
c
tur
e
,
r
e
c
or
ding
a
va
lue
o
f
0
.
9952
a
nd
a
low
M
S
E
o
f
0.
777
.
T
he
s
e
c
ond
tes
t
with
R
M
S
s
hows
r
e
s
ult
s
ne
a
r
ly
identica
l
to
C
NN
1,
with
a
n
a
c
c
ur
a
c
y
of
0.
8358
a
nd
C
NN
2
a
t
0.
8357;
howe
ve
r
,
the
M
S
E
va
lues
dif
f
e
r
s
igni
f
ica
ntl
y,
w
i
th
C
NN
1
a
t
19
.
142
a
nd
C
N
N2
a
t
41.
971
.
I
n
the
th
ir
d
tes
t,
us
ing
M
AV
,
notabl
e
r
e
s
ult
s
we
r
e
a
c
hieve
d,
with
C
NN
1
a
tt
a
ini
ng
a
n
R
2
s
c
or
e
o
f
0.
9177
a
nd
C
NN
2
a
n
R
2
s
c
or
e
o
f
0
.
8418.
F
igur
e
8
de
s
c
r
ibes
the
pe
r
f
o
r
manc
e
c
ompar
is
on
of
C
NN
a
r
c
hit
e
c
tur
e
ba
s
e
d
on
input
f
e
a
tur
e
s
.
F
igur
e
8(
a
)
de
pict
a
c
ompar
is
on
c
ha
r
t
of
the
R
2
a
nd
F
igur
e
8(
b)
de
pict
a
c
ompar
is
on
c
ha
r
t
of
the
R
M
S
E
of
a
r
c
hit
e
c
tur
e
s
1
a
nd
2
,
ba
s
e
d
on
th
r
e
e
input
f
e
a
tur
e
s
.
C
NN
1
a
r
c
hit
e
c
tur
e
pe
r
f
o
r
ms
s
igni
f
ica
ntl
y
be
tt
e
r
with
r
a
w
da
ta
than
R
M
S
or
M
AV
,
s
ugge
s
ti
ng
that
r
a
w
da
ta
c
ontains
mor
e
r
e
leva
nt
inf
o
r
mation
f
or
pr
e
c
is
e
pr
e
dictions
.
How
e
ve
r
,
r
e
a
l
-
ti
me
im
pleme
ntation
on
r
obots
m
us
t
c
ons
ider
pr
oc
e
s
s
ing
powe
r
a
nd
memor
y
c
ons
tr
a
int
s
,
r
e
quir
ing
a
ba
lanc
e
be
twe
e
n
a
c
c
ur
a
c
y
a
nd
pr
a
c
ti
c
a
li
ty.
F
utu
r
e
r
e
s
e
a
r
c
h
c
ould
e
xp
lor
e
a
lt
e
r
na
ti
ve
or
c
ombi
ne
d
input
f
e
a
tur
e
s
to
e
nha
nc
e
C
NN
pe
r
f
or
manc
e
.
Ove
r
a
ll
,
a
r
c
hit
e
c
tur
e
1
with
r
a
w
da
ta
a
ppe
a
r
s
to
be
the
opti
mal
c
hoice
f
or
r
e
a
l
-
ti
me
r
oboti
c
a
ppli
c
a
ti
ons
.
T
a
ble
2.
C
NN
a
r
c
hit
e
c
tur
e
tes
ti
ng
r
e
s
ult
s
S
ubj
e
c
t
R
A
W
R
M
S
M
A
V
R
2
M
S
E
R
2
M
S
E
R
2
M
S
E
C
N
N
1
C
N
N
2
C
N
N
1
C
N
N
2
C
N
N
1
C
N
N
2
C
N
N
1
C
N
N
2
C
N
N
1
C
N
N
2
C
N
N
1
C
N
N
2
1
0.9967
0.9964
0.640
0.880
0
.
8376
0
.
8376
18
.
019
40
.
122
0
.
9202
0
.
8611
20
.
381
33
.
972
2
0.9949
0.9899
1,100
2,131
0
.
8165
0
.
8165
23
.
145
45
.
430
0
.
8800
0
.
8565
29
.
642
39
.
805
3
0.9933
0.9908
0.784
0.935
0
.
8184
0
.
8184
20
.
632
43
.
787
0
.
9338
0
.
8410
17
.
020
45
.
176
4
0.9967
0.9952
0.646
1.198
0
.
8217
0
.
8217
15
.
925
47
.
623
0
.
9201
0
.
8171
20
.
874
48
.
346
5
0.9948
0.9943
0.717
1.385
0
.
8846
0
.
8846
17
.
990
32
.
896
0
.
9345
0
.
8337
17
.
921
44
.
231
M
e
a
n
0.9952
0.9933
0.777
1.305
0.8358
0
.
8357
19
.
142
41
.
971
0.9177
0.8418
21.676
42.306
(
a
)
(
b)
F
igur
e
8.
C
ompar
is
on
of
a
ve
r
a
ge
pe
r
f
o
r
manc
e
be
t
we
e
n
a
r
c
hit
e
c
tur
e
1
a
nd
a
r
c
hit
e
c
tur
e
2:
(
a
)
R
2
a
nd
(
b)
M
S
E
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
Ar
ti
f
I
ntell
,
Vol.
14
,
No.
4
,
Augus
t
2025
:
322
8
-
3240
3234
3.
2.
Grip
s
t
r
e
n
gt
h
p
r
e
d
ict
ion
t
e
s
t
in
g
Af
ter
e
xpe
r
im
e
nti
ng
with
va
r
ious
types
of
a
r
c
hit
e
c
tur
e
,
the
pr
e
dictive
r
e
s
ult
s
of
gr
ip
s
tr
e
ngth
we
r
e
tes
ted
.
T
he
tes
ts
we
r
e
c
a
r
r
ied
out
on
the
input
da
ta
f
r
o
m
R
M
S
a
nd
M
AV
us
ing
the
f
i
r
s
t
C
NN
a
r
c
hit
e
c
tur
e
.
F
igur
e
9
pr
e
dict
the
r
e
s
ult
s
of
C
NN
model
in
F
igur
e
9(
a
)
us
ing
R
M
S
a
nd
F
igur
e
9(
b)
us
ing
M
AV
f
e
a
tur
e
s
c
a
n
f
oll
ow
the
tar
ge
t
but
p
r
oduc
e
s
os
c
il
lating
pr
e
dictions
.
F
or
ins
tanc
e
,
a
t
a
tar
ge
t
of
10
N
,
the
pr
e
dicte
d
output
f
luctua
tes
a
r
ound
thi
s
va
lue
r
a
ther
than
be
i
ng
e
xa
c
t.
S
moot
hing
tec
hniques
li
ke
movi
ng
a
ve
r
a
ge
s
c
a
n
im
pr
ove
r
e
a
l
-
ti
me
pe
r
f
o
r
man
c
e
.
W
hil
e
it
is
unc
lea
r
whe
ther
R
M
S
o
r
M
AV
pe
r
f
o
r
ms
be
tt
e
r
,
the
R
M
S
-
ba
s
e
d
model
e
xhibi
ted
f
e
we
r
os
c
il
lations
.
R
a
w
da
ta
wa
s
e
xc
luded
to
mi
nim
ize
p
r
oc
e
s
s
ing
de
mands
f
or
r
e
a
l
-
ti
me
c
ontr
ol.
Ove
r
a
ll
,
the
C
NN
a
r
c
hit
e
c
tur
e
e
f
f
e
c
ti
ve
ly
pr
e
dicts
gr
ip
s
tr
e
ngth,
thou
gh
m
inor
de
viations
f
r
o
m
tar
ge
t
va
lues
r
e
main.
(
a
)
(
b)
F
igur
e
9.
C
NN
pr
e
diction
r
e
s
ult
s
with
f
e
a
tur
e
e
xtr
a
c
ti
on
us
ing
(
a
)
R
M
S
a
nd
(
b
)
M
AV
3.
3.
CN
N
an
d
ot
h
e
r
m
e
t
h
o
d
s
F
igur
e
10
pr
e
s
e
nts
the
r
e
s
ult
s
of
a
c
ompar
is
on
of
t
he
C
NN
method
with
f
our
other
methods
,
na
mely
L
S
T
M
,
R
F
,
DT
,
a
nd
K
NN
.
F
igur
e
10(
a
)
il
lus
tr
a
ted
t
he
c
ompar
is
on
mea
s
ur
e
s
us
e
R
²
a
nd
F
igu
r
e
10(
b
)
il
lus
tr
a
ted
M
S
E
.
T
he
s
e
f
igur
e
s
pr
ovide
a
de
tailed
c
ompar
is
on
of
the
f
ive
di
f
f
e
r
e
nt
methods
.
F
igur
e
10
s
how
two
int
e
r
e
s
ti
ng
thi
ngs
to
dis
c
us
s
:
t
he
e
f
f
e
c
t
of
f
e
a
tur
e
s
on
model
pe
r
f
or
manc
e
a
nd
the
model's
pe
r
f
or
manc
e
.
I
n
te
r
ms
o
f
f
e
a
tur
e
s
,
the
R
M
S
f
e
a
tur
e
is
be
tt
e
r
tha
n
the
M
AV
in
a
ll
models
.
T
hus
,
the
R
M
S
f
e
a
tur
e
is
mos
t
r
e
c
omm
e
nde
d
c
ompar
e
d
to
the
M
AV
f
e
a
tur
e
.
As
f
or
model
pe
r
f
or
manc
e
,
de
e
p
lea
r
ning
models
a
r
e
g
e
ne
r
a
ll
y
be
tt
e
r
than
mac
hine
lea
r
ning
models
,
with
the
C
N
N
model
be
ing
the
be
s
t.
I
f
we
look
a
t
the
pe
r
f
or
manc
e
of
the
mo
de
ls
f
r
om
the
R
2
s
ide,
it
s
e
e
ms
that
the
dif
f
e
r
e
nc
e
in
the
pe
r
f
or
manc
e
of
the
C
NN
model
a
nd
the
other
models
is
not
too
s
igni
f
ica
nt.
How
e
ve
r
,
the
di
f
f
e
r
e
nc
e
is
quit
e
s
igni
f
ica
nt
whe
n
view
e
d
f
r
o
m
th
e
R
M
S
E
va
lue.
T
he
C
NN
model
with
the
R
M
S
f
e
a
tur
e
pr
oduc
e
s
a
n
R
M
S
E
e
r
r
or
of
a
bout
2
Ne
wtons
,
whe
r
e
a
s
the
L
S
T
M
is
a
r
ound
6
Ne
wtons
.
E
ve
n
in
the
DT
,
it
c
a
n
be
up
to
10
Ne
wtons
.
M
e
a
nwhile,
the
highes
t
f
or
c
e
va
lue
mea
s
ur
e
d
by
the
s
e
ns
or
is
50
Ne
wton.
T
he
r
e
f
o
r
e
,
t
he
R
M
S
E
e
r
r
or
va
lue
of
10
Ne
wtons
is
ve
r
y
lar
ge
.
W
it
h
a
n
R
M
S
E
e
r
r
or
va
lue
of
a
r
ound
2
Ne
wtons
,
the
pe
r
f
or
manc
e
of
the
C
NN
mode
will
look
ve
r
y
good
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
Ar
ti
f
I
ntell
I
S
S
N:
2252
-
8938
M
y
oe
lec
tr
ic
gr
ip
for
c
e
pr
e
diction
us
ing
de
e
p
lear
n
ing
for
hand
r
obot
(
K
hair
ul
A
nam
)
3235
(
a
)
(
b)
F
igur
e
10
.
C
ompar
is
on
be
twe
e
n
f
ive
methods
ba
s
e
d
on:
(
a
)
R
2
a
nd
(
b)
M
S
E
3.
4.
On
l
in
e
e
xp
e
r
im
e
n
t
s
T
he
pr
oc
e
s
s
of
tes
ti
ng
pr
e
dictions
dir
e
c
tl
y
be
f
or
e
t
he
y
a
r
e
a
ppli
e
d
to
the
ha
nd
r
obot
is
c
a
ll
e
d
onli
ne
tes
ti
ng.
T
his
tes
t
make
s
us
e
of
E
M
G
da
ta,
with
R
M
S
a
nd
M
AV
e
xtr
a
c
ti
on
f
e
a
tur
e
s
,
us
ing
the
C
NN
method
that
wa
s
pr
e
vious
ly
tr
a
ined
in
the
of
f
li
ne
e
xpe
r
im
e
nt.
F
igur
e
11
p
r
e
s
e
nt
th
e
r
e
s
ult
s
f
o
r
the
onl
ine
te
s
t
of
the
C
NN
method
us
ing
R
M
S
a
r
e
p
r
e
s
e
nted
in
F
i
gur
e
11(
a
)
a
nd
M
AV
e
xtr
a
c
ti
on
f
e
a
tur
e
s
a
s
s
hown
in
F
igur
e
11(
b)
a
r
e
ve
r
y
s
im
il
a
r
.
How
e
ve
r
,
upon
c
los
e
r
c
ompar
is
on,
the
onli
ne
tes
t
with
R
M
S
f
e
a
tur
e
e
xtr
a
c
ti
on
ha
s
a
pr
e
dicte
d
va
l
ue
that
is
a
lm
os
t
identica
l
to
the
a
c
tual
va
lue.
(a)
(b
)
F
igur
e
11
.
Online
tes
t
r
e
s
ult
s
with
(
a
)
R
M
S
a
nd
(
b)
M
AV
f
e
a
tur
e
e
xtr
a
c
ti
on
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
Ar
ti
f
I
ntell
,
Vol.
14
,
No.
4
,
Augus
t
2025
:
322
8
-
3240
3236
3.
5.
Robot
i
m
p
lem
e
n
t
a
t
ion
3.
5.
1.
A
c
t
u
a
t
or
t
e
s
t
in
g
S
e
r
vomot
or
a
c
tuator
tes
ti
ng
is
pe
r
f
o
r
med
to
c
a
li
br
a
te
the
a
ngle
of
r
otation
of
the
s
e
r
vomot
or
.
A
pa
r
a
ll
e
l
a
r
c
is
plac
e
d
a
t
a
n
a
ngle
of
0
to
obtain
a
n
a
ngle
that
c
o
r
r
e
s
ponds
to
the
a
ngula
r
pa
r
a
mete
r
s
f
or
e
a
c
h
s
tr
e
ngth.
T
he
a
ngle
mea
s
ur
e
ment
da
ta
f
r
om
the
s
e
r
vo
mot
or
is
then
a
ve
r
a
ge
d,
a
nd
the
pe
r
c
e
nt
e
r
r
or
va
lue
is
c
a
lcula
ted.
T
h
e
r
e
s
ul
t
in
g
m
e
a
s
u
r
e
m
e
n
t
o
f
the
a
n
gl
e
o
f
th
e
s
e
r
v
o
mo
to
r
in
th
e
ha
nd
r
ob
ot
is
p
r
e
s
e
nte
d
in
T
a
bl
e
3
.
T
a
ble
3
.
Ac
tuator
tes
ti
ng
F
or
c
e
(
N
)
A
ngl
e
s
(
o
)
S
e
r
vo
a
ngl
e
me
a
s
ur
e
me
nt
(
o
)
A
ve
r
a
ge
a
ngl
e
(
o
)
A
ve
r
a
ge
e
r
r
or
(
%
)
T
humb
I
nde
x f
in
ge
r
M
id
dl
e
f
in
ge
r
R
in
g f
in
ge
r
s
L
it
tl
e
f
in
ge
r
0
50
48
50
48
49
49
48.8
2.45
10
70
65
68
65
67
67
66.4
5.42
20
90
85
85
80
80
85
84
7.14
30
110
100
105
100
100
105
103
6.79
40
130
120
120
120
120
120
120
8.3
50
150
140
140
140
140
140
140
7.14
I
n
th
is
e
va
luation,
we
de
ter
mi
ne
d
the
a
c
c
ur
a
c
y
p
a
r
a
mete
r
of
the
s
e
r
vo
r
o
tation
with
a
maximum
e
r
r
or
va
lue
of
10%
.
T
he
tes
t
r
e
s
ult
s
r
e
ve
a
led
that
the
highes
t
pe
r
c
e
ntage
e
r
r
or
obs
e
r
ve
d
wa
s
8.
3%
.
Our
mea
s
ur
e
ments
of
the
a
ngles
on
the
r
obot's
f
inger
pr
oduc
e
d
s
li
ghtl
y
dif
f
e
r
e
nt
a
ve
r
a
ge
va
lues
f
or
e
a
c
h
a
ngle
mea
s
ur
e
ment
c
ompar
e
d
to
the
a
c
tual
a
ngle.
T
h
i
s
is
s
u
pp
o
r
t
e
d
by
t
he
pe
r
c
e
nt
a
g
e
o
f
e
r
r
o
r
v
a
l
ue
s
tha
t
f
a
l
l
be
low
1
0
%
.
As
s
uc
h
,
we
c
a
n
c
o
n
f
i
de
nt
ly
a
f
f
i
r
m
th
a
t
a
l
l
t
he
s
e
r
vo
r
ota
t
io
ns
o
f
e
a
c
h
f
in
ge
r
wo
r
k
wi
th
u
tm
o
s
t
a
c
c
u
r
a
c
y
.
3.
5.
2.
Han
d
gr
ip
s
t
r
e
n
gt
h
t
e
s
t
in
g
in
r
ob
ot
s
T
he
C
NN
p
r
e
diction
r
e
s
ult
s
a
r
e
s
e
nt
to
A
r
duino
to
dr
ive
the
s
tr
e
ngth
o
f
the
r
obot
g
r
ip.
T
he
r
e
s
ult
s
o
f
the
gr
ip
tes
t
on
the
r
obot
a
r
e
s
hown
in
F
igu
r
e
1
2.
F
or
the
p
r
e
dicte
d
s
tr
e
ngth
of
0
a
nd
10
N
,
t
he
ope
r
a
tor
s
hould
hold
the
c
up
be
c
a
us
e
it
is
not
s
tr
ong
e
nough
f
or
the
r
obot
to
gr
ip
the
plas
ti
c
c
up.
T
he
r
obot
c
a
n
gr
ip
plas
ti
c
c
ups
we
ll
f
or
g
r
ip
s
tr
e
ngth
gr
e
a
ter
than
20
N.
F
igur
e
12
.
R
obot
ha
nd's
gr
ip
s
tr
e
ngth
mea
s
ur
e
d
in
e
xpe
r
im
e
nt
3.
6.
Dis
c
u
s
s
ion
an
d
l
im
it
a
t
ion
T
he
inves
ti
ga
ti
on
of
gr
ip
s
tr
e
ngth
p
r
e
diction
thr
ough
a
dva
nc
e
d
C
NN
a
r
c
hit
e
c
tur
e
s
r
e
pr
e
s
e
nts
a
s
igni
f
ica
nt
a
dva
nc
e
ment
in
r
oboti
c
ha
nd
c
ont
r
ol
a
nd
b
iom
e
c
ha
nica
l
s
ignal
pr
oc
e
s
s
ing.
Our
pr
opos
e
d
methodology
de
mons
tr
a
tes
r
e
ma
r
ka
ble
pr
e
dictiv
e
pe
r
f
o
r
manc
e
,
a
c
hieving
a
n
R
²
s
c
or
e
of
0.
99
,
whic
h
s
ubs
tantially
outper
f
or
ms
pr
e
vious
methodologi
c
a
l
a
ppr
oa
c
he
s
in
the
f
ield
of
gr
ip
s
tr
e
ngth
e
s
ti
m
a
ti
on.
A
c
r
it
ica
l
e
xa
mi
na
ti
on
of
e
xis
ti
ng
r
e
s
e
a
r
c
h
r
e
ve
a
ls
a
pr
ogr
e
s
s
ive
im
pr
ove
ment
in
pr
e
dictive
a
c
c
ur
a
c
y
a
c
r
os
s
va
r
ious
mac
hine
-
lea
r
ning
tec
hniques
.
As
il
lus
tr
a
te
d
in
the
c
ompar
a
ti
ve
pe
r
f
o
r
manc
e
T
a
ble
4,
the
pe
r
f
or
manc
e
of
gr
ip
s
tr
e
ngth
pr
e
diction
methods
ha
s
e
volved
f
r
om
li
ne
a
r
r
e
gr
e
s
s
ion
(
R
²=
0.
82)
to
inc
r
e
a
s
ingl
y
s
ophis
ti
c
a
ted
ne
ur
a
l
ne
twor
k
a
ppr
oa
c
he
s
.
Our
C
NN
-
ba
s
e
d
method
r
e
pr
e
s
e
nt
s
the
c
ur
r
e
nt
pinnac
le
of
pe
r
f
or
manc
e
,
s
igni
f
ica
ntl
y
a
dva
nc
ing
the
s
tate
-
of
-
the
-
a
r
t
in
gr
ip
s
tr
e
ngth
pr
e
diction
.
T
a
ble
4.
C
ompar
a
ti
ve
pe
r
f
o
r
manc
e
of
gr
ip
s
tr
e
ngt
h
pr
e
diction
methods
R
e
s
e
a
r
c
h
m
e
th
ods
R
2
s
c
or
e
R
e
gr
e
s
s
io
n
[
24]
0.82
M
L
P
[
18]
0.88
L
S
T
M
[
25]
0.90
NN
[
26]
0.98
C
N
N
(
pr
opos
e
d me
th
od
)
0.99
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
Ar
ti
f
I
ntell
I
S
S
N:
2252
-
8938
M
y
oe
lec
tr
ic
gr
ip
for
c
e
pr
e
diction
us
ing
de
e
p
lear
n
ing
for
hand
r
obot
(
K
hair
ul
A
nam
)
3237
De
s
pit
e
the
pr
omi
s
ing
r
e
s
ult
s
,
s
e
ve
r
a
l
c
r
it
ica
l
li
mi
tations
mus
t
be
a
ddr
e
s
s
e
d
in
f
utur
e
r
e
s
e
a
r
c
h.
T
he
c
ur
r
e
nt
s
ys
tem's
pr
e
diction
r
e
s
ult
s
,
while
ge
ne
r
a
ll
y
a
c
c
ur
a
te,
e
xhibi
t
oc
c
a
s
ional
f
luctua
ti
ons
that
c
ould
pos
e
r
is
ks
whe
n
ha
ndli
ng
de
li
c
a
te
objec
ts
.
T
his
va
r
iabil
it
y
ne
c
e
s
s
it
a
tes
f
ur
ther
r
e
f
ineme
nt
to
e
ns
ur
e
c
ons
is
tent
a
nd
pr
e
c
is
e
gr
ip
s
tr
e
ngth
p
r
e
diction.
M
or
e
ove
r
,
the
s
tudy's
c
ur
r
e
nt
s
c
ope
is
li
mi
ted
to
tes
ti
ng
on
he
a
lt
hy
s
ubjec
ts
,
c
r
e
a
ti
ng
a
s
igni
f
ica
nt
r
e
s
e
a
r
c
h
ga
p
in
unde
r
s
tanding
the
s
ys
tem's
e
f
f
e
c
ti
v
e
ne
s
s
f
or
a
mput
e
e
pop
ulations
.
F
u
tu
r
e
s
t
ud
ies
s
ho
ul
d
p
r
io
r
it
iz
e
e
x
te
nd
in
g
the
r
e
s
e
a
r
c
h
to
di
ve
r
s
e
s
ub
je
c
t
g
r
o
ups
,
pa
r
t
ic
ul
a
r
l
y
i
nd
iv
i
du
a
ls
wi
th
l
i
mb
d
if
f
e
r
e
n
c
e
s
,
to
va
li
da
te
a
n
d
o
pt
im
iz
e
t
he
p
r
o
pos
e
d
me
th
od
ol
og
y
.
T
h
e
s
t
ud
y's
f
in
d
in
gs
e
xt
e
n
d
be
yon
d
me
r
e
t
e
c
hn
ica
l
a
c
hi
e
ve
me
nt
,
o
f
f
e
r
in
g
p
r
o
f
ou
nd
i
ns
i
gh
ts
in
t
o
th
e
po
ten
t
ia
l
of
d
e
e
p
lea
r
ni
ng
m
e
t
ho
ds
f
o
r
gr
ip
s
t
r
e
ng
th
p
r
e
d
ic
ti
on
.
T
he
p
r
o
pos
e
d
C
NN
a
r
c
hi
te
c
t
ur
e
n
ot
o
n
ly
d
e
m
ons
t
r
a
tes
s
u
pe
r
io
r
p
r
e
di
c
t
iv
e
c
a
pa
b
i
li
ti
e
s
b
ut
a
ls
o
op
e
ns
n
e
w
a
ve
n
ue
s
f
o
r
a
dva
nc
e
d
p
r
os
t
he
ti
c
c
on
t
r
o
l
,
r
e
ha
b
i
li
ta
ti
on
tec
hn
ol
og
ies
,
a
nd
h
u
man
-
r
ob
ot
i
nte
r
a
c
t
io
n
i
n
te
r
f
a
c
e
s
.
B
y
b
r
i
dg
in
g
th
e
ga
p
b
e
t
we
e
n
bi
om
e
c
ha
n
ic
a
l
s
i
gn
a
l
p
r
oc
e
s
s
in
g
a
n
d
mac
h
ine
l
e
a
r
n
in
g
,
t
his
r
e
s
e
a
r
c
h
c
on
t
r
i
bu
tes
t
o
th
e
b
r
oa
de
r
s
c
ie
nt
i
f
i
c
un
de
r
s
t
a
n
di
ng
o
f
p
r
e
c
is
e
f
o
r
c
e
c
on
t
r
o
l
in
r
o
bo
ti
c
a
nd
a
s
s
is
t
iv
e
tec
hn
o
lo
gi
e
s
.
F
u
tu
r
e
r
e
s
e
a
r
c
h
s
ho
ul
d
f
oc
u
s
on
r
e
f
i
ni
ng
the
p
r
e
d
ic
t
ive
mo
de
l
,
e
xp
l
or
in
g
t
r
a
ns
f
e
r
le
a
r
n
in
g
c
a
p
a
b
il
i
t
ies
,
a
nd
c
on
du
c
t
in
g
e
xt
e
ns
iv
e
r
e
a
l
-
wo
r
ld
va
li
da
ti
on
t
r
i
a
ls
to
u
n
loc
k
the
f
ul
l
p
o
ten
t
ial
o
f
t
h
is
i
nn
ova
t
iv
e
a
p
p
r
oa
c
h
.
4.
CO
NC
L
USI
ON
T
he
a
im
of
the
s
tudy
wa
s
to
a
s
s
e
s
s
the
e
f
f
e
c
ti
ve
ne
s
s
of
C
NN
in
c
ontr
oll
ing
the
gr
ip
s
tr
e
ngth
o
f
a
r
oboti
c
ha
nd
by
pr
e
dicting
us
e
r
gr
ip
s
tr
e
ngth
thr
ou
gh
E
M
G
s
ignals
.
E
M
G
s
ignals
a
r
e
ge
ne
r
a
ted
whe
n
mus
c
les
c
ontr
a
c
t,
a
nd
they
c
a
n
p
r
ovide
a
n
a
c
c
ur
a
te
mea
s
u
r
e
ment
of
g
r
ip
s
tr
e
ngth.
T
he
s
tudy
e
va
luate
d
two
dif
f
e
r
e
nt
C
NN
a
r
c
hit
e
c
tur
e
s
,
C
NN
1
a
nd
C
NN
2,
to
de
ter
mi
n
e
the
be
s
t
a
ppr
oa
c
h.
C
NN
1
wa
s
de
s
igned
with
e
ight
de
pth
laye
r
s
,
whil
e
C
NN
2
ha
d
s
ix
de
pth
laye
r
s
a
nd
uti
li
z
e
d
r
a
w
a
nd
R
M
S
input
da
ta.
Af
ter
a
na
lyzing
the
r
e
s
ult
s
,
C
NN
1
pr
ove
d
to
be
the
s
upe
r
ior
a
r
c
hit
e
c
tu
r
e
.
T
he
pr
e
dicte
d
gr
ip
s
tr
e
ngth
wa
s
s
uc
c
e
s
s
f
ull
y
tr
a
ns
mi
tt
e
d
to
the
r
oboti
c
ha
nd,
but
the
s
ys
tem
did
not
maintain
a
c
ons
i
s
tent
leve
l
of
s
tr
e
ngth.
T
he
r
e
f
or
e
,
f
utur
e
r
e
s
e
a
r
c
h
will
f
oc
us
on
a
ddr
e
s
s
ing
thi
s
is
s
ue
to
im
pr
ove
the
s
y
s
tem's
pe
r
f
or
manc
e
.
T
he
s
tudy
highl
ight
s
the
pot
e
nti
a
l
of
de
e
p
lea
r
ning
tec
hniques
,
s
uc
h
a
s
C
NN
,
in
c
ontr
ol
li
ng
r
oboti
c
ha
nd
gr
ip
s
tr
e
ngth
.
W
it
h
f
ur
the
r
a
dva
n
c
e
ments
in
thi
s
a
r
e
a
,
th
is
tec
hnology
c
a
n
ha
ve
a
s
igni
f
ica
nt
im
pa
c
t
on
im
pr
oving
the
qua
li
ty
of
li
f
e
f
or
ind
ivi
d
ua
ls
with
li
mi
ted
ha
nd
f
unc
ti
ona
li
ty.
F
UN
DI
NG
I
NF
ORM
AT
I
ON
T
his
r
e
s
e
a
r
c
h
is
f
inanc
ially
s
uppor
ted
by
Gr
a
nt
of
I
nter
na
ti
ona
l
C
oll
a
bor
a
ti
o
n
R
e
s
e
a
r
c
h
2022,
I
ns
ti
tut
e
f
or
R
e
s
e
a
r
c
h
a
nd
C
omm
unit
y
S
e
r
vice
,
Unive
r
s
it
y
of
J
e
mber
,
unde
r
c
ontr
a
c
t
number
4393/UN25.
3.
1/L
T
/2022
.
AU
T
HO
R
CONT
RI
B
U
T
I
ONS
S
T
AT
E
M
E
N
T
T
his
jour
na
l
us
e
s
the
C
ontr
ibut
o
r
R
oles
T
a
xo
nomy
(
C
R
e
diT
)
to
r
e
c
ognize
indi
vidual
a
uthor
c
ont
r
ibut
ions
,
r
e
duc
e
a
utho
r
s
hip
dis
putes
,
a
nd
f
a
c
il
it
a
te
c
oll
a
bor
a
ti
on.
Nam
e
of
Au
t
h
or
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
Kha
ir
ul
Ana
m
✓
✓
✓
✓
✓
✓
✓
Dhe
ny
Dw
i
Ar
dhians
ya
h
✓
✓
✓
✓
✓
✓
✓
M
uc
ha
mad
Ar
if
Ha
na
S
a
s
ono
✓
✓
✓
✓
Ar
iza
l
M
uji
btama
la
Na
nda
I
mr
on
✓
✓
✓
✓
✓
Na
uf
a
l
Ainur
R
iza
l
✓
✓
✓
✓
M
oc
ha
mad
E
dowa
r
d
R
a
madha
n
✓
✓
✓
✓
Ar
is
Z
a
inul
M
utt
a
qin
✓
✓
✓
✓
C
laudio
C
a
s
telli
ni
✓
✓
✓
S
umar
di
✓
✓
✓
✓
C
:
C
onc
e
pt
ua
li
z
a
ti
on
M
:
M
e
th
odol
ogy
So
:
So
f
twa
r
e
Va
:
Va
li
da
ti
on
Fo
:
Fo
r
ma
l
a
na
ly
s
is
I
:
I
nve
s
ti
ga
ti
on
R
:
R
e
s
our
c
e
s
D
:
D
a
ta
C
ur
a
ti
on
O
:
W
r
it
in
g
-
O
r
ig
in
a
l
D
r
a
f
t
E
:
W
r
it
in
g
-
R
e
vi
e
w
&
E
di
ti
ng
Vi
:
Vi
s
ua
li
z
a
ti
on
Su
:
Su
pe
r
vi
s
io
n
P
:
P
r
oj
e
c
t
a
dmi
ni
s
tr
a
ti
on
Fu
:
Fu
ndi
ng a
c
qui
s
it
io
n
Evaluation Warning : The document was created with Spire.PDF for Python.