Indonesi
an
Journa
l
of El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
13
,
No.
2
,
Febr
uar
y
201
9
, pp.
5
27
~
533
IS
S
N: 25
02
-
4752, DO
I: 10
.11
591/ijeecs
.v1
3
.i
2
.pp
5
27
-
533
527
Journ
al h
om
e
page
:
http:
//
ia
es
core.c
om/j
ourn
als/i
ndex.
ph
p/ij
eecs
Optimi
zation
of
video qu
ality vi
a fuzzy l
ogic base
adapti
ve media p
layout
Farij
E
ht
ib
a
1
, B
ad
li
shah
Ahm
ad
2
,
Mos
tafi
ju
r
Rahm
an
3
1
Depa
rtment of
Com
m
unic
at
ion and
Com
pute
r
Networks,
Inforrm
at
ion
Technol
og
y
S
chool
,
Misurata Unive
rsit
y
,
L
ibay
2
ENAC Re
sea
r
c
h,
School
of
Co
m
pute
r
and
Com
m
unic
at
ion
Enginee
ring
,
Univ
ersity
Ma
lay
sia
Per
li
s,
Ma
lay
sia
3
Depa
rtment of
Software
Eng
ineeri
ng,
Daffodi
l
I
nte
rna
ti
ona
l
Uni
ver
sit
y
(DIU
), Dhaka
,
Bang
la
des
h
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Sep
9
, 2
018
Re
vised
N
ov
10
, 2
018
Accepte
d
Nov
2
4
, 201
8
Thi
s
pape
r
addr
esses
the
issue
of
int
err
upti
on
of
Vira
ble
Bit
Rat
e
(VBR)
vide
o
str
ea
m
in
g
over
IP
n
etw
ork,
due
to
cha
nne
l
qua
li
t
y
fluc
tu
ation.
Speci
fi
ca
l
l
y
,
a
Fuzz
y
Logic
(F
L)
Control
pr
in
ci
pl
e
is
fused
w
it
h
Adapt
ive
Media
Pla
y
ou
t
(
AM
P)
cont
rol
to
esti
m
at
e
an
app
ropria
t
e
play
ou
t
fra
m
e
rate,
namel
y
FLAM
P.
Based
on
t
he
esti
m
ation
of
the
pl
a
y
out
fra
m
e
rate,
the
FLAM
P
al
g
orit
hm
adj
usts
t
he
consum
pti
on
rat
e
sm
oothly
to
avoi
d
video
strea
m
degr
ada
t
ion.
Sim
ula
ti
on
result
s
val
ida
t
e
tha
t
the
FLA
MP
sche
m
e
eff
icientl
y
red
u
c
es
the
buff
er
ou
ta
ge
prob
abi
l
ity
and
provide
s
b
et
t
er
visual
qual
ity
wh
ere
F
LAMP
give
s
26.
5%
le
ss
fer
qu
e
nce
of
pl
a
y
out
i
nte
rrupt
ion,
and
21.
6%
l
ess v
ari
an
ce of
distor
t
ion
of
p
lay
out
as
compare
d
to
AP
TA.
Ke
yw
or
ds:
Ad
a
ptive m
edia p
la
yo
ut
Fu
zzy
lo
gic
Vide
o
stream
ing
Vira
ble b
it
ra
te
Copyright
©
201
9
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed.
Corres
pond
in
g
Aut
h
or
:
Farij Ehti
ba
,
Dep
a
rtm
ent o
f C
omm
un
ic
at
io
n
a
nd Com
pu
te
r
N
et
w
orks,
Inforr
m
at
ion
T
echnolo
gy Sc
hool,
Mi
su
rata
Un
i
ve
rsity
, Lib
ay
.
Em
a
il
: F.eh
ti
ba@
it
.m
isurr
at
a
u.
e
du.ly
1.
INTROD
U
CTION
Vide
o
c
onfer
e
nce,
vid
e
o
on
Dem
and
(C
oD)
a
nd
I
nter
ne
t
Pr
ot
oco
l
Te
le
vision
(IPT
V)
a
re
good
exam
ples
of
m
ultim
edia
that
ref
le
ct
s
a
bo
om
of
m
edia
strea
m
ing
se
rv
ic
e
s
over
the
I
nte
rn
et
[1
]
.
I
n
ge
ner
al
,
vid
e
o
stream
in
g
is
transm
i
tt
e
d
to
the
cl
ie
n
t
in
a
con
ti
nuou
s
m
ann
er,
wh
i
ch
can
be
wat
ched
im
m
edatl
y
as
i
t
receive
d.
H
owever,
the
deliv
ery
of
vid
e
o
s
tream
ing
over
IP
netw
orks
sti
ll
faces
m
any
chall
eng
es
due
to
var
ia
ble b
a
ndw
idth
wh
ic
h
cas
ues dela
y jit
te
r, an
d packet l
oss [
2,
3].
Com
m
on
ly
,
data
buff
e
rin
g
is
placed
at
the
cl
ie
nt
to
al
le
vi
at
e
the
de
grad
at
ion
of
net
work
qual
it
y.
Pr
act
ic
al
ly
,
so
m
e
vid
eo
fr
am
es
are
store
d
in
the
receive
d
buff
e
r
to
reli
ev
e
the
play
ou
t
interr
upti
ons
re
su
lt
ing
from
var
ia
ble
band
width.
T
hus,
the
receive
r
buff
e
r
is
m
on
it
or
e
d
f
o
r
t
rig
ger
i
ng
the
a
da
ptive
of
play
out
rat
e
.
Aggressi
vely
a
doptio
n
te
ch
ni
qu
e
,
w
hich
co
ns
ist
s
of
sim
pl
y
discard
i
ng
or
du
plica
te
d
vi
deo
fr
am
e,
is
us
e
d
to
adjust
the
play
ou
t
f
ram
e
rate.
Eve
n
thou
gh
the
pro
ba
bili
ty
of
play
ou
t
de
gr
a
datio
n
dec
r
eases
as
m
or
e
vid
e
o
fr
am
es
are
bu
f
fer
e
d,
the
play
ou
t
delay
incre
ases
[4
-
6].
O
n
the
oth
e
r
ha
nd,
A
ggressi
vely
adoptio
n
can
c
ause
play
ou
t
disru
ption
s
or
disco
nt
inu
it
ie
s w
it
h a
conseq
ue
nt d
e
gr
a
datio
n of
t
he
Quali
ty
o
f Ex
per
ie
nce (Q
oE) [
7].
AMP
re
duces
play
ou
t
delay
by
m
ini
m
iz
e
th
e
stora
ge
data
and
a
djust
s
pla
yout
r
rate
base
d
on
c
hannel
qu
al
it
y
[
8].
A
MP
te
ch
niques
are
sm
oo
th
a
dju
stm
ent
te
ch
niques
base
d
on
c
ha
ng
i
ng
the
play
out
r
at
e
within
acce
ptable
ra
nges.
I
n
ge
ner
al
,
the
ob
j
ect
ive
of
the
e
xisti
ng
AMP
te
c
hn
i
qu
e
s
is
t
o
a
da
pt
the
pla
yo
ut
rate
to
com
pen
sat
e
th
e
de
viati
on
bet
ween
the
ar
riv
al
rate
a
nd
co
nsum
ption
rate.
Howe
ver,
m
os
t
of
the
e
xisti
ng
AM
P
te
chn
iq
ues
a
re
i
m
ple
m
ented
by
two
m
ai
n
ste
ps
,
trig
geri
ng
play
out
ra
te
and
inter
pola
ti
on
play
out
rate.
Re
gardin
g
to
the
inter
po
la
ti
on
play
out
rate,
lots
AMP
te
chn
i
qu
e
s
are
ba
sed
on
li
near
pl
ay
ou
t
ad
j
us
tm
ent
[9
-
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
13
, N
o.
2
,
Fe
bru
ary 2
019
:
5
2
7
–
5
3
3
528
12
]
.
I
n
c
on
t
rast,
a
qu
a
drat
ic
pl
ay
ou
t
a
dju
stm
e
nt
is
porpose
d
in
[
13]
w
hich
giv
es
bette
r
pe
rfor
m
ance
in
te
rm
of
reducin
g
t
he dist
or
ti
on
of the
play
ou
t c
urve
.
Since
the
pla
yout
cu
r
ve
s
m
oo
thn
ess
is
a
m
ai
n
key
of
sat
isfyi
ng
th
e
QoE,
a
ne
w
te
ch
nique,
nam
ely
FLA
MP,
is
pr
op
os
e
d
in
this
pap
er
to
inv
est
igate
and
dem
on
stra
te
that
AMP
based
on
Fu
zzy
log
ic
con
t
ro
l
giv
es
a
good
perf
or
m
ance
in
te
rm
s
of
m
ini
m
iz
ing
the
probabil
it
y
of
buf
fer
un
derflo
w
,
t
he
ra
ti
o
of
buff
e
r o
verflo
w,
t
he variat
io
n of play
out
of d
ist
ort
ion an
d abr
upt va
riat
io
ns
i
n
the
p
la
yo
ut cur
ve.
2.
DESIG
N OF
FLAM
P TE
C
HNIQUE
2
.
1.
S
ystem
Model
Figure
1
il
lustr
at
es
the
arch
it
ect
ur
e
of
ty
pic
al
vid
eo
st
rea
m
ing
desi
gn
over
IP
ne
tw
ork
s.
The
vid
e
o
stream
o
rigin
at
ing
from
a strea
m
ing
serve
r
r
eaches t
he
cl
ie
nt,
a
nd que
ued
to b
e
d
is
play
ed
.
Figure
1.
Ty
pi
cal
arch
it
ect
ure
of F
L
AMP tec
hn
i
qu
e
v
i
deo st
ream
ing
over I
P
net
wor
k
The
cl
ie
nt
has
the
abili
ty
to
m
anipu
la
te
the
fr
am
e
du
rati
on
d
∈
S={
d1,
d2,
…,dn
},
w
her
e
d1
>
d2,>
⋯
>
dn
a
nd
dn
is
the
norm
al
vid
eo
fr
am
e
separ
at
ion.
H
ow
e
ve
r,
The
ti
m
e
o
f
play
out
is
div
ide
d
into
ti
m
e
slots
wh
ic
h
hav
e
sa
m
e
tim
e
du
rati
on
s
of
t=
dn.
Wh
e
n
t
he
vi
de
o
f
ram
es
reach
es
the
cl
ie
nt’s
buff
e
r
a
re
que
ued
f
or
disp
la
yi
ng
la
te
r.
Let
An
∈
A={
0,1,…,A
n}
de
no
te
s
the
num
ber
of
ar
rival
f
ram
e
du
ri
ng
ti
m
e
slot
t,
wh
er
e
An
is
the m
axi
m
u
m
nu
m
ber
of ar
rival fram
es.
Since
the
cha
nnel
qu
al
it
y
is
fluctuated
a
nd
t
he
inter
ar
rival
s
of
vi
deo
fr
a
m
es
are
var
ia
bl
e,
the
fr
am
e
arr
ival
process
m
a
y
be
f
ur
th
er
ra
ndom
iz
e.
Let
D
n
∈
D={
1,2,..,
D
n
}
de
note
the
num
ber
of
f
ram
es
disp
la
ye
d
durin
g
slot
t.
Defi
ne
f
ur
t
herm
or
e
B
LVL
as
t
he
buf
fer
fu
ll
ne
ss
le
vel
in
te
r
m
s
of
the
num
ber
of
fr
am
es
at
the
beg
i
nn
i
ng of sl
ot t. T
he
n
the
next b
uffe
r
f
ulln
ess level B
LVL+1
an
d
it
s
de
viat
ion
is
cate
goriz
ed
as:
n
n
L
V
L
L
V
L
D
A
B
B
1
(1)
1
L
V
L
L
V
L
VAR
B
B
B
(2)
The
fr
am
e
pla
yout
durati
on
(
d)
is
c
on
st
rain
ed,
w
hich
re
presents
t
he
ti
m
e
inter
val
to
be
increase
d
or
decr
ease
d
in
a
sing
le
ste
p
acc
ordin
g
to
c
hannel
qual
it
y.
Su
bs
e
qu
e
ntly
,
the
con
ce
pt
of
FL
con
tr
ol
is
fuse
d
wit
h
AMP
to
tu
ne
t
he
inte
rv
al
ti
m
e
(P
OUT
)
a
nd
a
dap
ts
play
out
rate
accu
ratel
y.
Th
e
goal
of
FLA
MP
desi
gn
is
to
m
anipu
la
te
the
f
ram
e p
la
yout
to co
m
pen
sat
e
the d
e
viati
on
be
tween c
ons
um
pt
ion
rate an
d
a
rr
ival
rat
e.
2
.
2
.
S
ystem
Model
The
a
rch
it
ect
ure
of
FL
AMP
con
t
ro
l
te
ch
ni
qu
e
is
present
ed
in
Fig
ur
e
2.
It
co
ntains
of
two
i
nputs,
buff
e
r
f
ullness
(B
LVL
)
a
nd
vari
ance
of
buf
fer
f
ullness
(B
VAR
),
a
re
c
hose
n
a
s
in
pu
ts
f
or
th
e
FL
co
ntr
ol
t
o
t
une
the
play
out
ra
te
.
Firstl
y,
the
input
var
ia
bles
are
c
onve
rt
ed
int
o
s
uitab
le
li
ng
uisti
c
va
riables.
Seco
nd
ly
,
the
li
ng
uisti
c
var
ia
bles
are
e
m
ulate
d
base
d
on
the
knowle
dge
base
to
m
ake
accurate
decisi
on
s.
Finall
y,
the
res
ult
of
F
uzzy
Infer
e
nce
Syst
e
m
(F
IS
)
is
def
uzzi
fi
ed
into
a
crisp
va
lue
w
hich
re
pr
ese
nts
the
a
ccur
at
e
interval ti
m
e (P
OUT
)
to
disp
la
y t
he
ne
xt
vid
e
o fr
am
e.
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Op
ti
miz
atio
n o
f vi
deo
qualit
y
vi
a
fuzzy
lo
gic
ba
s
e
adapti
ve med
i
a play
ou
t
(
Farij Ehtib
a
)
529
Figure
2. Ef
fec
ts of sel
ect
ing
diff
e
re
nt sw
it
c
hing
unde
r dyn
a
m
ic
co
nd
it
io
n
2
.
2
.
1.
Fuz
z
y
Sets
Members
h
ip
of FL
AM
P
Te
chnique
A
fu
zzy
set is def
i
n
ed
as a g
r
oup
of ele
m
ent
s w
it
h
a con
ti
nuum
o
f
m
e
m
ber
sh
i
p
gr
a
des
. Each
ele
m
ent
has
a
gr
a
de
of
m
e
m
ber
sh
ip
r
ang
i
ng
betwee
n
[0
-
1].
In
FL
AMP
te
chn
i
que,
two
input
pa
ram
et
ers
,
Figu
res
3
and
4,
an
d
an
ou
t
pu
t
pa
ram
eter
,
Fig
ur
es
5,
are
util
iz
ed
to
regulat
e
the
play
ou
t
rate.
H
oweve
r,
the
in
put
and
ou
t
pu
t
pa
ram
eter
s a
re
us
ed
to defi
ne
the
fi
ve l
ing
uisti
c
var
ia
bles as
sho
wn in Ta
ble
1
.
Table
1.
Inp
uts
and
Ou
t
pu
t
Linguist
ic
V
a
riab
le
s
of FL
AMP
Fu
zzy
Sets
Mem
ber
sh
ip
Variable
Fu
zzy
Se
t
M
e
m
b
e
r
sh
ip
B
LVL
[V
-
Low,
Low, No
r
m
a
l,
High
,
V
-
Hig
h
]
B
VAR
[N
-
Hig
h
,
N
-
Low,
No
r
m
al
,
P
-
Low,
P
-
Hig
h
]
P
OUT
[V
-
Low,
Low, No
r
m
a
l,
High
,
V
-
Hig
h
]
Figure
3. The
m
e
m
ber
sh
ip
f
unct
ion o
f b
uffe
r
f
ullness
level
(B
LVL
)
Figure
4.
The
m
e
m
ber
sh
ip
f
unct
ion o
f
buf
fe
r
f
ullness
level
(B
VAR
)
R
ec
eiv
er
B
u
f
f
er
µ
B
L
V
L
B
V
A
R
P
OUT
F
u
zzi
f
icati
o
n
P
r
o
ce
s
s
F
I
S
Def
u
zzif
icati
o
n
P
r
o
ce
s
s
F
u
z
z
y
s
e
ts
Ru
le Bas
e
K
n
o
w
led
g
e
B
a
se
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
13
, N
o.
2
,
Fe
bru
ary 2
019
:
5
2
7
–
5
3
3
530
Figure
5.
The
m
e
m
ber
sh
ip
f
unct
ion o
f play
out rate
(P
OUT
)
2
.
2
.
2.
T
he FL
AMP Te
chniq
ue Rul
e
-
Ba
se
The
FLA
M
P
r
ule
-
base
c
onsist
s
of
gro
up
of
f
uzzy
ru
le
s
.
They
us
e
t
o
f
or
m
ulate
the
c
onditi
on
al
sta
tem
ents
that
com
pr
ise
fu
zzy
log
ic
.
The
phil
osophy
of
the
FLA
M
P
te
chn
i
qu
e
r
ule
is
to
adap
t
the
pl
ay
ou
t
rate
w
hen
the
dev
ia
ti
on
betw
een
ar
rival
rate
an
d
co
nsum
ing
rate
occ
ur
s
.
Table
2
il
lustra
te
s
the
IF
-
T
he
n
r
ules
of
the
FL
AMP
te
chn
iq
ue
w
hich
the
colum
n
represe
nts
the
fu
zzy
set
of
bu
ff
e
r
fu
ll
ne
ss
an
d
the
row
re
present
s
the fuzzy
set
of the
var
ia
ti
on
of
buff
e
r fu
ll
nes
s.
Table
2.
T
he
Rule
-
Ba
se
of FL
AMP Tec
hniq
ue
Variables
Bu
ff
er
Variation
(
B
VAR
)
Neg
ativ
e High
Neg
ativ
e L
o
w
No
r
m
al
Po
sitiv
e L
o
w
Po
sitiv
e High
Bu
ff
er
Fu
lln
ess
Level
(B
LVL
)
Ver
y
L
o
w
Ver
y
L
o
w
Ver
y
L
o
w
Ver
y
L
o
w
Ver
y
L
o
w
Ver
y
L
o
w
Low
Ver
y
L
o
w
Low
Low
Low
No
r
m
al
No
r
m
al
Low
No
r
m
al
No
r
m
al
No
r
m
al
Hig
h
Hig
h
No
r
m
al
Hig
h
Hig
h
Hig
h
Ver
y
High
Ver
y
High
Ver
y
High
Ver
y
High
Ver
y
High
Ver
y
High
Ver
y
High
2
.
2
.
3.
Fuz
z
y
Inf
erence
Proc
ess (
FI
P)
of F
LAMP
FI
P,
th
e
kernel
of
a
fu
zzy
lo
gic
syst
e
m
,
is
the
process
of
form
ulati
ng
the
m
app
ing
fro
m
the
giv
en
inputs
to
an
ou
t
pu
t
us
i
ng
fu
zzy
l
og
ic
.
Since
Ma
m
dan
i
-
based
m
odel
is
a
c
omm
on
ly
us
ed
a
s
f
uzzy
m
et
ho
dolo
gy,
i
t
has
bee
n
c
hosen
as
the
m
et
hodolo
gy
of
t
he
FL
AMP
c
on
t
ro
ll
er
[
14
]
.
Th
e
FL
AMP
te
ch
nique
involves
f
our
s
ta
ges
[15].
Th
e
FI
S
em
plo
ys
bu
f
fe
r
fu
ll
ne
s
s
and
it
s
var
ia
t
ion
to
tu
ne
the
per
fect
play
out
rate
as foll
ows.
The
fi
rst
ste
p
is
to
determ
i
ne
the
m
e
m
be
rsh
i
p
de
gree
of
the
i
nput
va
riables
(B
LVL
and
B
VAR
),
wh
e
re
the
de
viati
on
of
buff
e
r
fu
ll
nes
s
is
cal
culat
ed
as
s
ho
wn
(
2)
:
T
he
se
cond
ste
p
is
to
evaluate
the
F
LAMP
ru
le
s
of
the.
T
he
degree
of
m
e
m
ber
sh
ip
of
each
pr
ec
edi
ng
in
put
va
riabl
e
is
assesse
d
usi
ng
t
he
ru
le
s
-
base
o
f
the
FLA
M
P
te
chn
i
qu
e
.
T
he
thir
d
ste
p
is
to
aggre
gate
the
r
ulers
’
outp
uts.
The
m
e
m
ber
shi
p
value
s
of
al
l
ru
le
s
are c
om
po
und
into a
n ou
t
pu
t
value.
The
f
ourt
h
ste
p
is
the
De
fu
zi
ficat
ion
.
T
he
a
ggre
gate
outp
ut
fu
zzy
set
is
defuzzifi
ed
t
o
ob
t
ai
n
a
sin
gle
nu
m
ber
w
hich
rep
re
sents
the
new
play
out
inter
val
(P
OUT
).
The
center
of
gr
a
vity
(COG)
m
e
tho
d
is
ch
os
e
n
defuzific
at
io
n process
. For
m
al
ly
, th
e CO
G
is
calc
ulate
d
as
foll
ows.
b
a
S
S
F
b
a
S
S
S
F
C
O
G
)
(
)
(
(3)
3.
E
X
PERI
MEN
TS A
ND R
E
S
ULTS
This
sect
io
n
c
har
act
erizes
a
nd
com
par
es
th
e
perform
ance
of
t
he
FL
AM
P
te
chn
i
qu
e
w
it
h
pr
e
vious
te
chn
iq
ues
.
T
he
sim
ulatio
n
s
et
ti
ng
s
a
nd
th
e
vi
deo
strea
m
are
descr
i
be
d.
To
e
valuat
e
the
pe
rfor
m
ance
of
con
t
ro
l
te
c
hn
i
ques,
the
m
et
rics
are
desi
gn
at
e
d.
Finall
y,
the
si
m
ulati
on
res
ults
of
al
l
te
ch
nique
a
re
disc
us
se
d
and com
par
e
d.
Fo
r
the
FL
AM
P tech
nique
va
li
dation,
va
rio
us AMP
tech
niques
ha
ve been
exam
ined.
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Op
ti
miz
atio
n o
f vi
deo
qualit
y
vi
a
fuzzy
lo
gic
ba
s
e
adapti
ve med
i
a play
ou
t
(
Farij Ehtib
a
)
531
a)
No Co
ntr
ol: t
he
v
ide
o st
ream
is d
isplay
ed
b
a
ck wit
hout a
da
ptit
ion
c
on
t
ro
l.
b)
Linear
Sl
owd
own
a
nd
Sp
ee
dup:
T
he
play
ou
t
adap
ta
ti
on
c
ontr
ol,
w
hich
s
pe
edup
a
nd
sl
ow
dow
n
play
out
rate, is
perform
ed
to
adj
us
t
play
ou
t
rate acc
or
ding
to
thr
e
sho
ld level
[8
]
.
c)
AP
T
A
te
ch
nique:
The
AP
T
A
al
go
rit
hm
is
e
m
plo
ye
d
to
va
li
date
the
propose
d
te
chn
i
que
s.
The
AP
T
A
al
gorithm
is
a
thres
ho
l
d
base
te
chn
iq
ue
w
hic
h
the
play
out
r
at
e
is
quad
rati
cal
ly
decr
ease
d
wh
e
n
t
he
data
le
vel
is
belo
w
low
t
hr
es
hold.
In
c
ontrast
,
it
i
s
quad
rati
cal
ly
increase
d
wh
e
n
the
data
le
ve
l
exceeds
hi
gh
thres
ho
l
d.
Ot
he
rw
ise
,
it
is
adap
te
d
de
pe
nd
i
ng
on
the
inst
ant
arr
ival
rate
[1
3].
F
ur
t
herm
or
e,
the
AP
T
A
par
am
et
ers
are
set
as in [
13
]
.
3.1.
Si
mul
at
io
n S
et
tin
g
Netw
ork
Sim
ulator
2
(
NS2
)
is
adopted
t
o
validat
e
the
FLA
MP
te
ch
nique
[
14
]
.
A
ccordin
gly,
the
MPE
G
-
4
e
ncode
d
vid
e
o,
Ju
ras
sic
Par
k
I
[15]
is
ob
ta
in
ed
as
a
vi
de
o
stream
so
ur
ce.
The
vi
deo
pa
cke
t
s
wh
ic
h
a
re
se
nt
over
IP
net
w
ork
to
t
he
cl
ie
nt
is
j
it
te
re
d
with
diff
e
re
nt
patt
ern
s
.
Si
nce
Pa
r
et
o
O
N
-
O
FF
s
ource
can
j
it
te
r
vi
de
o
stream
m
or
e
sig
nificantl
y
than
ot
her
s
,
a
P
areto
with
pa
r
a
m
et
ers
(40
0
m
s,
60
0
m
s)
ha
s
bee
n
chosen
to
j
it
te
r
the
vid
e
o
st
re
a
m
[9
]
.
T
he
a
r
rival
ti
m
e
of
al
l
pack
et
s
are
store
d
a
nd
us
ed
as
the
a
rr
i
val
proces
s
of the
AMP tec
hn
i
qu
e
s.
3.2.
Perf
orm
ance
M
e
trics
The
var
ia
ti
on
of
play
bac
k
s
peed
af
fects
the
vid
e
o
strea
m
ing
play
out,
wh
ic
h
ca
us
es
degra
dation
of
vid
e
o qu
al
it
y.
Ther
e
f
or
e,
se
ve
ra
l ef
fecti
ve m
et
rics fo
r
AM
P tech
niques
are defi
ned as
the foll
owin
g.
3.2.
1.
Freque
n
cy of
Pla
yout
Int
erru
pt
i
on
s
The
vi
deo
str
ea
m
is
interr
up
te
d
w
hen
t
her
e
is
no
fra
m
e
in
the
c
li
ent’s
buff
e
r.
Natur
al
ly
,
le
ss f
re
quency
of p
la
yo
ut
inte
rrup
ti
ons
is b
et
te
r
pe
rfor
m
anc
e.
T
U
N
U
f
(4)
wh
e
re
re
pr
ese
nts the
num
ber
of
play
ou
t i
nterrup
ti
ons
dur
i
ng
play
the
vid
e
o
cl
ip ti
m
e T.
3.2.
2.
V
arian
c
e of
Dist
ort
io
n
of Pla
you
t
(
V
DoP)
Play
ou
t
inter
r
up
ti
on
str
ongl
y
aff
ect
s
the
qu
al
it
y
of
e
xperience
(
Q
oE).
The
disto
rtio
n
of
play
out
(DoP
)
is
pr
es
e
nted
i
n
[13]
to
stud
y t
he
ef
fec
ts of play
o
ut in
te
rr
upti
on
on vi
deo
stream
ing
.
2
1
2
D
o
P
N
N
i
i
D
o
P
V
D
o
P
(5)
wh
e
re
N rep
res
ents total
re
cei
ved f
ram
es includ
in
g discar
de
d fr
am
es.
3.2.
3
.
Pla
yout
Cu
r
ve
Play
ou
t
cu
r
ve
represe
nt
the
num
ber
of
byte
s
hav
e
bee
n
pl
ay
ed
by
the
cl
ie
nt
durin
g
the
interval
of
tim
e
[
13
]
.
T
hus,
the
play
out
interr
up
ti
on
af
fects
the
sm
oo
thn
ess
of
vid
e
o
play
bac
k.
I
n
gen
e
ral,
the
pl
ay
ou
t
curve
cl
arifies
the
per
f
or
m
ance
o
f
an
AMP
te
chn
iq
ue
w
hich
sm
oo
th
play
ou
t
curve
in
di
cat
es
a
go
od
AM
P
te
chn
iq
ue pe
rfor
m
ance.
3.3.
Sim
ulat
i
on
Resul
ts
This
stu
dy
in
ve
sti
gates
the
play
ou
t
inte
rrup
t
ion
,
va
riance
of
dist
or
ti
on
of
play
ou
t
a
nd
th
e
cum
ulati
ve
play
ou
t
data
f
or
va
rio
us
A
MP
te
ch
niqu
e
s.
Fi
gure
6
s
hows
the
f
requ
ency
of
play
out
inter
r
up
ti
on
s
f
or
al
l
play
ou
t
c
ontr
ol
te
chn
i
qu
e
s.
T
he
f
re
qu
e
ncy
of
play
ou
t
i
nterrup
ti
ons
substa
ntial
ly
increase
d,
es
pecial
ly
at
high
cro
ss
tr
af
fic
load
val
ues.
Howev
e
r,
t
he
res
ul
ts
of
the
FL
A
MP
ou
t
weig
hs
the
oth
e
r
existi
ng
te
c
hn
i
qu
es
i
n
te
r
m
of
t
he
fr
e
que
ncy
of
play
ou
t
interr
up
ti
on.
It
giv
es
a
r
ound
53.
4%
co
m
par
ing
to
N
o
C
on
tr
ol
te
c
hn
i
qu
e
,
and 21.
6%
com
par
ing
to
AP
TA
te
c
hn
i
qu
e
.
Figure
7
dem
on
st
rates
the
var
ia
nce
of
di
stortio
n
of
pla
yout
f
or
al
l
AMP.
Since
the
fu
zzy
log
ic
i
m
i
ta
te
s
hu
m
an
br
ai
n,
the
pl
ay
ou
t
rate
ad
aptat
ion
is
perfect
ly
tun
e
d.
Ther
e
f
or
e,
T
he
FL
AMP
te
c
hn
i
qu
e
m
anipu
la
te
s
th
e
play
ou
t
rate
eff
ic
ie
ntly
ba
sed
on
netw
ork
qual
it
y.
The
FLA
MP
resul
t
giv
es
26.
5%
le
ss
distor
ti
on
of
play
ou
t c
om
par
ing t
o AP
TA
and
46% c
om
par
ing t
o N
o
C
on
tr
ol.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
13
, N
o.
2
,
Fe
bru
ary 2
019
:
5
2
7
–
5
3
3
532
Figure
8
il
lus
trat
es
to
the
play
ou
t
cu
r
ves
of
AMP
te
chn
i
qu
e
s.
As
m
entioned
befor
e
,
the
low
perver
sio
n
is
the
m
os
t
accura
te
te
chn
iq
ue.
T
he
play
out
c
urve
of
F
LAMP
te
chn
iq
ue
is
th
e
lowest
as
c
om
par
ed
to
e
xisti
ng
te
c
hn
i
qu
e
s.
Co
nse
qu
e
ntly
,
this
r
esult
ind
ic
at
es
the
reli
abili
ty
of
pr
e
vious
res
ults,
the
inter
r
up
ti
on
of the
vid
e
o pl
ay
ou
t
resu
lt
a
nd the
v
is
ual
qual
it
y deg
ra
dation res
ult.
Figure
6
.
The
f
reque
ncy of
play
ou
t i
nte
rru
ption
f
or
AMP
te
chn
i
qu
e
s
Figure
7
.
The
var
ia
ti
on
of
dis
tortio
n of
play
ou
t
f
or
AMP
te
chn
i
qu
e
s
Figure
8
.
The
play
ou
t c
urves
for AM
P
te
c
hniqu
es
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Op
ti
miz
atio
n o
f vi
deo
qualit
y
vi
a
fuzzy
lo
gic
ba
s
e
adapti
ve med
i
a play
ou
t
(
Farij Ehtib
a
)
533
4.
CONCL
US
I
O
N
In
t
his
stu
dy,
a
new
AMP
te
chn
iq
ue
ba
sed
on
t
he
F
L
co
ntro
l
is
pro
po
se
d
a
nd
evaluate
d.
The
te
ch
nique,
FLA
MP, esti
m
at
es the p
la
yo
ut r
at
e b
a
sed on
the
netw
ork
q
ualit
y. More
ov
e
r,
the F
L c
oncept is
fu
se
d
wit
h
the
AMP
te
ch
niqu
e
to
trigg
e
r
an
d
interp
olate
play
ou
t
rate.
Nu
m
erical
resu
lt
s
in
dicat
e
that
FLA
MP
te
chn
iq
ue
giv
e
s
bette
r
pe
rfo
r
m
ance
in
te
r
m
of
reducin
g
play
ou
t
inter
rupt
ion
,
vis
ual
qua
li
ty
deg
rad
at
io
n
an
d
sm
oo
thn
ess
of
play
ou
t.
REFERE
NCE
S
[1]
Al
-
Zoubi
,
H., et
al
.
,
As
urve
y
On
Rec
en
t
Advan
ces
in
IPTV.
JJ
CIT,
2
,
(2):
p.
86
-
10
6.
(2016)
.
[2]
W
u,
J.,
et
al.,
Co
nte
nt
-
aware
con
current
multi
pat
h
transfer
for
high
-
def
inition
v
id
eo
streaming
ov
er
het
eroge
n
eou
s
wirel
ess networ
k
s.
TPDS
,
27,
(3):
p.
710
-
723.
(20
16).
[3]
Su,
G.
-
M.,
et
al
.
,
QoE
in
vi
d
eo
streaming
ove
r
w
irel
ess
net
works
:
perspec
tives
a
nd
research
cha
ll
eng
es.
W
N,
22
,
(5):
p.
1571
-
159
3.
(2016)
.
[4]
Su,
G.
-
M
.
,
e
t
al.
,
QoE
in
v
ide
o
streaming
ove
r
wirel
ess
net
wor
ks:
perspec
tives
and
research
ch
all
eng
es.
QoE
i
n
vide
o
str
ea
m
ing over
wire
le
ss
ne
t
works
:
per
spec
t
i
ves
and resea
r
ch
challe
ng
es:
p
.
1
-
23.
(2015)
.
[5]
Zha
ng,
X.
and
H.
Hass
ane
in,
A
sur
ve
y
of
p
ee
r
-
to
-
pee
r
live
vi
d
e
o
streaming
sche
mes
–
An
algor
it
hmic
p
erspec
tive
.
Com
pute
r
Netw
orks,
56,
(15):
p.
3548
-
3579.
(20
12).
[6]
Li
ndeb
erg
,
M.
,
et
a
l.,
Chall
enges
and
te
chni
qu
e
s
for
vi
deo
stre
aming
ove
r
mobile
ad
ho
c
ne
tw
orks.
Multi
m
edia
S
y
stems
,
17,
(1):
p.
51
-
82.
(2011)
.
[7]
Hoßfel
d,
T.,
e
t
a
l.
Quanti
fi
ca
ti
on
of
Y
ouTube
QoE
vi
a
crowdsour
ci
ng
.
in
2011
I
E
EE
Int
ernati
ona
l
Symposium
on
Mult
imedi
a
(
ISM)
,
.
IEEE. (201
1).
[8]
Kalman,
M.,
E
.
Stei
nba
ch,
and
B.
Girod,
Adapti
ve
media
pla
you
t
for
low
-
del
ay
v
ide
o
streaming
ove
r
error
-
pron
e
channe
ls.
TCSVT,
14
,
(6)
:
p
.
841
-
851.
(2004)
.
[9]
Chen,
Y.
and
G.
Li
u
.
Adap
ti
v
e
media
pla
yout
a
ss
iste
d
rate
ada
ptat
ion
sch
eme
f
or
HT
TP
adapti
ve
streaming
ov
er
lt
e
system
.
in
Mu
lt
imedi
a
&
Ex
po
Workshops
(
IC
MEW
)
,
2016
IE
EE
In
te
rnationa
l
Confe
ren
ce on
.
IEE
E
.
(2016)
.
[10]
Yang,
J.,
e
t
al.
,
Online
buf
fe
r
ful
lne
ss
esti
ma
ti
on
aide
d
ada
pti
v
e
media
pla
yout
for
vi
d
eo
streaming.
IEEE
Tra
nsac
ti
ons on
Multi
m
edi
a
,
13
,
(5):
p.
1141
-
115
3.
(2011)
.
[11]
Hu,
H.,
et
al
.
,
Sc
ene
aware
sm
oo
th
play
out
cont
r
ol
for
portable
media
play
ers
ove
r
random
VB
R
ch
annel
s.
T
-
C
E,
56,
(4
):
p
.
2330
-
2338.
(2010)
.
[12]
Su,
Y.
-
F.,
et
al.,
Smooth
cont
rol
of
adapti
ve
med
i
a
play
out
for
vi
d
eo
streaming.
IE
EE
Tra
nsa
ct
ions
on
Multi
m
edi
a,
11,
(7):
p.
1331
-
1339.
(2009)
.
[13]
Li
,
M
.
,
T.
-
W
.
Lin,
and
S.
-
H.
Che
ng,
Arrival
proc
ess
-
cont
rolle
d
a
dapti
v
e
media
p
l
ayout
wi
th
multiple
thresholds
fo
r
vi
deo
streaming.
Multi
m
edia
S
y
s
te
m
s,
18,
(5):
p.
391
-
407.
(2012)
.
[14]
The
Net
work
Si
mulator
-
ns
-
2
.
Avail
ab
le
f
rom
: http:
/
/www
.
isi.
e
du/nsnam/ns/.
[15]
MPE
G
-
4
and
H.263
Vi
d
eo
Tr
ace
s
for
Net
work
Pe
rform
ance
E
val
uati
on
.
Avai
l
abl
e
from
:
ht
tp://www
2.
tkn.
tu
-
ber
li
n
.
de/
r
ese
ar
c
h/t
ra
ce
/
trace
.
h
tml.
Evaluation Warning : The document was created with Spire.PDF for Python.