Indonesi
an
Journa
l
of El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
10
,
No.
1
,
A
pr
il
201
8
, p
p.
146
~
153
IS
S
N: 25
02
-
4752, DO
I: 10
.11
591/ijeecs
.v1
0.i
1.
pp
146
-
1
53
146
Journ
al h
om
e
page
:
http:
//
ia
es
core.c
om/j
ourn
als/i
ndex.
ph
p/ij
eecs
Perform
ance E
valuati
on
o
f Mu
ltichannel
Audi
o
Comp
r
essi
on
Te
ddy
Sur
ya
Gun
awan
1
,
Mi
ra
K
art
iw
i
2
1
Depa
rtment of
El
e
ct
ri
ca
l
and
C
om
pute
r
Engi
n
e
eri
ng,
Kulliyy
a
h
of
Engi
n
ee
rin
g
,
I
ndia
2
Depa
rtment of
Inform
at
ion
S
y
s
t
ems
,
Kulliyy
ah
of
ICT
Inte
rna
ti
ona
l
Isl
a
m
ic
Univer
sit
y
Malay
s
ia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Ja
n
21
, 2
01
8
Re
vised
Ma
r
18
, 2
01
8
Accepte
d
Ma
r
29
, 201
8
In
re
c
ent
y
e
ars,
m
ult
ic
hann
el
au
dio
s
y
st
ems
are
widely
used
in
m
oder
n
sound
devi
c
es
as
it
can
provide
m
ore
re
al
isti
c
and
e
ngagi
ng
expe
r
ience
to
the
li
stener.
Thi
s
pa
per
foc
uses
on
the
per
form
anc
e
eva
lu
at
ion
of
thr
ee
loss
y
,
i.e
.
AA
C,
Ogg
Vor
bis,
and
Opus
,
a
nd
three
lossle
ss
compress
ion,
i.
e
.
FLAC,
True
Audio,
and
W
avPac
k,
for
m
ult
ic
h
annel
audio
signal
s,
in
cl
ud
ing
stereo
,
5.
1
and
7
.
1
channel
s.
Expe
rime
nts
were
conducte
d
on
the
sam
e
three
aud
io
fil
es
but
with
diffe
re
n
t
cha
nn
el
conf
igura
t
ions.
The
per
form
anc
e
of
ea
c
h
enc
oder
was
ev
a
lua
t
ed
base
d
on
i
ts
enc
od
ing
t
ime
(a
ver
age
d
over
100
ti
m
es)
,
dat
a
re
du
ct
ion
,
and
audi
o
qual
ity
.
Us
ual
l
y
,
the
re
is
al
wa
y
s
a
tra
de
-
off
bet
wee
n
th
e
thr
ee
m
et
ric
s
.
To
sim
pli
f
y
the
ev
al
ua
ti
on,
a
n
ew
int
egr
a
te
d
per
form
anc
e
m
e
tri
c
was
propos
e
d
that
combine
s
al
l
th
e
three
p
erf
orm
anc
e
m
et
ric
s.
Us
ing
the
new
m
ea
sure,
FLAC
was
fo
und
to
be
the
best
lossless
compress
ion,
wh
il
e
Ogg
Vorbis
a
nd
Opus
were
fo
und
to
b
e
the
be
st
for
los
s
y
compress
ion
depe
nds
on
the
cha
n
nel
conf
iguratio
n.
Thi
s
r
esult
co
uld
be
use
d
in
de
te
rm
ini
ng
t
he
prope
r
aud
io form
at
for
m
ulti
cha
nne
l audi
o
s
ystems
.
Ke
yw
or
d
s
:
M
ulti
channel
A
udio
L
os
sle
ss
C
om
pr
essio
n
L
os
sy
C
om
pr
ession
E
nc
od
i
ng
T
im
e
C
om
pr
ession
R
at
io
PEAQ
Copyright
©
201
8
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
:
Ted
dy S
ur
ya
G
un
a
wa
n
,
Dep
a
rtm
ent o
f El
ect
rical
an
d
Com
pu
te
r
E
ng
i
neer
i
ng, Kulli
yy
ah
of E
nginee
rin
g
,
In
te
r
natio
nal Is
lam
ic
U
niv
er
sit
y M
al
ay
sia
,
Jal
an Go
m
bak
,
5310
0 Ku
al
a
Lum
pu
r, (+
603)
6196
4521
.
Em
a
il
:
tsgu
na
wan@ii
um
.ed
u.m
y, tsguna
wa
n@gm
ai
l.co
m
1.
INTROD
U
CTION
In
rece
nt
ye
ars
,
m
ulti
chan
nel
aud
i
o
syst
e
m
s
are
widely
us
e
d
in
m
od
er
n
sound
de
vices.
U
su
al
ly
,
two
dig
it
s
separa
te
d
by
a
deci
m
a
l
po
int,
e.
g.
2.1,
4.1
,
5.1,
6.
1,
7.1
,
are
us
e
d
to
cl
assify
t
he
var
i
ous
kinds
of
sp
ea
ker
set
-
up
[1
]
,
[
2]
.
T
his
nu
m
ber
re
pres
ents
the
num
ber
of
au
dio
tra
cks
use
d.
S
ome
aud
i
o
syst
em
s
on
ly
hav
e
a
sin
gle
channel
or
tw
o
cha
nnel
s
(
ste
r
eoph
on
ic
sou
nd
or
2.0
c
hann
el
sound
).
The
first
di
git
sho
ws
t
he
nu
m
ber
of
pri
m
ary
chan
nels
,
i.e.
sat
el
li
te
un
it
s,
each
of
w
hich
are
repr
oduce
d
on
a
sin
gle
sp
ea
ker
whic
h
has
the
capa
bili
ty
t
o
ha
ndle
range
of
f
re
qu
e
ncy
betwee
n
10
0Hz
to
22
kHz.
On
t
he
oth
e
r
ha
nd,
the
sec
on
d
di
git
(d
eci
m
al
dig
it
)
rep
re
sents
the
pr
ese
nce
of
L
FE
(L
ow
F
re
quency
E
ff
ect
)
t
hat
is
rep
r
oduc
ed
on
a
sub
woof
e
r.
More
ov
e
r,
s
urrou
nd
syst
em
descr
i
bes
a
typ
e
of
a
ud
i
o
outp
ut
in
wh
ic
h
the
so
und
a
ppears
to
s
urround
the
li
ste
ner
by
360
de
gr
ees
,
in
w
hich
it
giv
e
s
im
pr
ession
t
hat
sound
a
re
c
om
ing
f
ro
m
al
l
po
s
sible
direct
ion
s
.
I
t
has bee
n used t
o pro
vid
e a
m
or
e reali
sti
c an
d eng
a
ging e
xper
ie
nce
[
3]
.
Ther
e
are
t
wo
kinds
of
a
udio
com
pr
essi
on
al
gorithm
tho
se
are
los
sy
and
l
os
sle
ss.
L
os
sy
au
di
o
com
pr
essio
n
is
known
by
the
ir
well
-
desi
gn
e
d
syst
e
m
to
sh
rin
ks
file
siz
es.
Adva
nced
A
ud
i
o
Co
ding
(
AA
C
)
,
MPEG
-
1
Lay
e
r
III
(MP
3),
D
olb
y
AC
-
3,
O
pu
s
,
O
G
G
V
orbis
[
4]
and
Wi
ndows
Me
dia
Audio
L
os
sy
(
WM
A
lossy)
are
the
exam
ples
of
popula
r
lossy
aud
i
o
co
ding
syst
e
m
[5]
.
AA
C
can
be
c
onside
red
as
th
e
m
os
t
influ
e
ntial
m
ulti
chan
nel
a
ud
i
o
co
ding
al
gori
thm
[6]
.
This
i
s
du
e
to
it
s
abi
li
ty
to
su
pport
aud
i
o
cha
nnel
s
up
t
o
48
cha
nnel
s
a
nd
co
ntribute
l
os
sle
ss
a
ud
i
o
f
or
5.1
c
ha
nn
el
s
at
sam
pling
rates
32
0
kbit
s
/s.
Me
an
wh
il
e,
AC3
pro
vid
es
hi
gh
aud
i
o
qual
it
y
a
t
38
4k
bit/
s
[7]
.
Me
anwhil
e,
the
m
os
t
well
-
known
co
dec
in
l
os
sle
ss
al
gorit
hm
are
Fr
ee
L
os
sle
ss
Audio
C
od
ec
(F
L
AC),
A
pp
l
e
Loss
le
ss
Audio
Co
dec
(ALAC)
,
W
a
vP
a
ck
(
WAV
),
M
PEG
-
4
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
Perf
orma
nce E
valu
ation of
M
ulti
chann
el
Audio
C
ompre
ssion (
Teddy
Sury
a Gun
awan)
147
Audio
lossless
[
8]
,
Tr
ue
A
udio
(T
TA
)
[9]
,
and
IE
EE
1857.
2
[10]
.
L
os
sl
ess
com
pr
essi
on
al
go
rithm
s
do
not
hav
e
an
y l
os
s i
nfor
m
at
ion
a
nd provide
an exa
ct
r
eplic
a
of th
e original si
gn
al
.
Althou
gh
m
any
research
has
been
c
onduct
ed
on
lossless
and
los
sy
aud
io
com
pr
essio
n,
but
no
t
m
any
researc
hes
ha
ve
been
f
oc
us
e
d
on
the
pe
rfo
rm
ance
evaluati
on
on
m
ulti
c
hannel
au
dio
c
od
i
ng.
The
refo
re,
the
ob
j
ect
ive
of
th
is
pap
e
r
is
to
inv
e
sti
gate
the
perform
ance
of
va
rio
us
au
dio
com
pr
ession
a
lgorit
hm
s
to
e
nc
od
e
m
ul
ti
chan
nel
a
ud
i
o
in
te
rm
s
of
enc
od
i
ng
ti
m
e,
data
savi
ng,
and
qual
it
y.
Furtherm
or
e,
a
ne
w
integ
rated
m
et
ric
was pr
opose
d t
o
inte
gr
at
e all
t
hr
ee
m
et
ri
cs.
2.
MU
LT
I
CHA
NN
EL
A
UD
I
O CON
FIG
U
RA
TI
ON
The
detai
ls
of
m
ul
ti
chan
nel
a
ud
i
o
s
pea
ker
c
onfig
ur
at
io
n
ha
s
bee
n
presen
te
d
in
[
11
]
.
F
r
om
analo
g
aud
i
o,
sam
pling
a
nd
qua
ntiza
ti
on
a
re
c
ondu
ct
ed
to
re
pr
ese
nt
the
sou
nd
wav
e
int
o
dig
i
ta
l
representat
i
on.
A
ste
reo
si
gn
al
c
an
be
co
ns
ide
r
ed
as
tw
o
in
de
pende
nt
cha
nn
el
s
of
a
udio
in
f
or
m
at
ion
,
i.e.
l
eft
an
d
ri
gh
t
c
ha
nn
el
s
.
Stereo
phonic
aud
i
o
pro
vid
es
the
im
pr
essio
n
of
s
ound
lo
c
al
iz
at
ion
.
Un
l
ike
m
on
o
an
d
ste
re
o
a
ud
i
o,
m
ul
ti
-
channel
a
ud
i
o
form
at
design
a
te
s
in
m
or
e
th
an
tw
o
c
hanne
ls.
This
ty
pe
of
au
dio
f
or
m
at
aim
s
to
ad
vance
the
abili
ty
of
so
und
local
iz
at
io
n.
As
a
n
exa
m
ple,
a
5.
1
m
ul
ti
chan
nel
louds
pea
ker
s
a
rr
a
ng
em
ent
ha
s
been
il
lus
trat
ed
in
F
ig
ure
1(a).
T
he
le
ft
and
rig
ht
channels
place
d
at
±30˚
li
ke
i
n
ste
re
o
au
dio
.
Me
anwhil
e,
th
e
rear
rig
ht
and
le
ft
channel
locat
e
d
at
±110˚.
U
su
al
ly
,
they
are
us
ed
f
or
ext
end
e
d
s
ound
so
urce
local
iz
at
ion
s
interp
retat
ion.
Fo
r
ce
nter
c
ha
nn
el
,
0˚
c
omm
o
nly
fo
r
pla
yi
ng
agai
n
voi
ce
con
te
nts
in
m
ov
in
g
au
di
o.
T
he
decim
al
dig
it
(.
1)
c
hannel
refe
r
to
sub
woo
fe
r
c
hannel
wh
ic
h
al
s
o
r
eco
gniz
e
as
L
FE
c
ha
nn
el
.
T
his
c
ha
nn
el
is
for
play
ing
ba
ck
the
lo
w
f
re
qu
e
ncy
c
on
te
nt
s.
By
ad
ding
m
or
e
su
r
rou
nd
louds
peaker
t
o
the
tw
o
sta
ndar
d
channels L
S a
nd RS,
it
w
il
l cr
eat
e large
r
li
ste
ning z
on
e
. T
hi
s setu
p had bee
n widely
used
in
ci
nem
a
[12]
.
Figure
1. 5.1
a
nd 7.1 M
ulti
ch
ann
el
Sp
ea
ke
rs
Setu
p
Mult
ic
hannel
aud
i
o
7.1
is
a
fu
rt
her
en
ha
nce
m
ent
to
5.
1
au
dio
cha
nnel
s.
Ther
e
a
re
oth
e
r
two
side
-
su
r
rou
nd
s
peaker
in
the
sp
e
ak
er
c
onfig
ur
at
io
n.
Ma
ny
of
ap
plica
ti
on
us
e
d
7.1
a
ud
i
o
i
n
order
to
great
er
i
m
pact
of
s
urr
ound
sound.
T
he
lo
ud
sp
ea
ker
s
a
rr
a
ngem
ent
is
almo
st
sim
il
ar
to
5.1
m
ulti
chan
nel
au
dio
.
Howev
e
r,
there
are
an
ot
he
r
two
s
peak
e
r
le
ft
and
rig
ht
rear
w
hich
a
bout
±13
5˚
to
s
urr
ound
sound.
Fi
g
ure
1(b
)
sho
w
s
the
set
up
c
onfig
ur
at
ion
of
m
ultich
an
nel
7.1
a
ud
i
o,
wh
il
e
T
able
1
s
hows
the
sta
ndar
d
channel
la
yo
uts
f
or
m
ul
ti
chan
nel
a
ud
i
o.
Be
yo
nd
7.1
m
ulti
chan
ne
l
aud
io
,
10.2
channel
s
urrou
nd
sou
nd
has
been
de
velo
pe
d.
It
is
the
ad
va
nced
ver
si
on
of
5.1
te
chnolo
gy,
but
10.
2
c
ould
pro
du
ce
twic
e
as
go
od
as
5.1.
In
this
c
ha
nn
el
config
ur
at
io
n.
14
c
hannels
ar
e
us
ed
to
in
cl
udin
g
five
fron
t
sp
eake
rs,
five
su
r
rou
nd
c
ha
nn
el
s
,
two
L
F
E
and
two heig
hts,
p
l
us
t
he
a
dd
it
io
n of a sec
ond su
b
-
woo
fer
[
12
]
.
Table
1.
Stan
da
rd Cha
nn
el
L
ay
ou
ts
Ch
an
n
el Na
m
e
Deco
m
p
o
sitio
n
Mon
o
FC
Stereo
FL
+
FR
2
.1
FL
+
FR +
LFE
5
.1
FL
+
FR +
FC +
L
FE
+ B
L
+
BR
7
.1
FL
+
FR +
FC +
L
FE
+ B
L
+
BR +
S
L
+ SR
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.
10
, N
o.
1
,
A
pr
il
2018
:
146
–
153
148
FL = Fr
on
t
Lef
t, FR =
Fron
t
Ri
gh
t,
FC = F
r
on
t C
e
nter, LF
E = L
ow F
requ
ency
BL = B
ack
Lef
t, BR
= Back
Ri
gh
t,
SL =
Side Le
ft, S
R
=
Side Rig
ht
3.
MU
LT
I
CHA
NN
EL
A
UD
I
O COM
PRES
SION AL
GO
RITH
MS
In
this
pa
per,
three
los
sy
a
nd
th
ree
l
os
sl
ess
a
ud
i
o
c
om
pr
ession
al
gorithm
s
will
be
e
valuate
d,
includi
ng
Advan
ce
d
A
udio
Cod
in
g
(
AAC
),
O
gg
Vor
bis,
O
pus,
F
LAC,
Tr
ue
Audio,
a
nd
Wa
vP
ac
k,
resp
ect
ively
.
No
te
that,
the
sel
e
ct
ed
cod
e
r
s
are
capab
le
t
o
ha
nd
le
m
ultich
an
nel
com
pressi
on
for
ste
r
eo,
5.1
,
and 7.1
ch
a
nne
ls.
3.1.
Advance
d
Audi
o Codi
ng (
A
AC)
AA
C
le
ads
M
P3
as
there
is
a
new
non
-
bac
kw
a
r
d
com
patib
le
aud
i
o
co
de
r
introd
uced
in
[1
]
,
[
6]
.
It
beco
m
es
popul
ar d
ue
to
ap
plica
ti
on
i
n
Appl
e
iTu
nes.
A
AC
operates
MDC
T
tra
ns
f
or
m
only
in
it
s
m
ai
n
cod
i
ng
loop
an
d
transi
ent
detect
ion
f
un
ct
io
n
to
dete
ct
a
lon
g
wind
ow
of
20
48
points
or
a
serial
set
of
ei
gh
t
25
6
point
windows
is
rea
dy
for
the
MD
CT
transfo
rm
.
Th
us
,
this
giv
e
high
fr
e
quen
c
y
reso
luti
on
of
23Hz
an
d
2.7
m
s
fo
r
a
sig
nal
sam
pl
ed
at
48
kHz.
A
gain
c
ontr
ol
proce
dure
is
inco
r
porated
i
n
the
S
SR
pr
of
i
le
of
A
AC.
A
Pseud
o
Qu
a
drat
ur
e
Mi
rr
or
Fil
te
r
(PQMF)
filt
er
ba
nk
is
use
d
t
o
s
plit
the
sig
nal
into
f
our
su
bba
nds
wi
th
sam
e
band
width.
T
he
or
i
gin
al
si
gnal
sam
pling
rates
re
du
ce
d
t
o
quarter
s
by
disc
ard
i
ng
one
or
m
or
e
su
bba
nd
s
.
A
AC
util
iz
es
the
te
m
po
ral
-
noise
-
s
harpin
g
te
ch
ni
qu
e
t
o
e
xp
el
t
he
pre
-
ech
o
e
f
fect
cause
d
by
transie
nts.
Ba
sed
on
su
bject
ive
ev
al
uations,
A
AC
pro
vid
e
s
great
aud
i
o for
5
c
ha
nn
el
ba
ndwidt
h
at
bit rate
of
320kbps.
3.2.
Ogg
Vo
r
bis
Ogg
Vor
bis
is a
fu
ll
o
pe
n
s
ource,
non
-
pro
pri
et
ary,
patent
a
nd
roya
lt
y
fr
ee
com
pr
essio
n
a
ud
i
o
f
orm
at
.
It
is
based
on
ve
ct
or
qu
a
ntiza
ti
on
a
nd
tr
ans
form
ation
with
overla
pp
i
ng
windows,
i
.e
.
m
od
ifie
d
disc
rete
cosi
ne
trans
form
(MDCT).
Each
wi
ndows
can
hav
e
2048
or
512
s
a
m
ples.
The
shorter
on
e
is
use
d
only
to
enco
de
a
transient
sig
nal
s.
Af
te
r
tra
nsf
or
m
at
ion
to
fr
e
qu
e
ncy
dom
ain
,
the
sig
nal
is
analy
zed
by
psy
cho
ac
ousti
c
m
od
e
l
and ina
ud
i
ble
par
t
of the
sp
ec
trum
is rem
ov
ed.
The
n
t
he
fl
oor
v
ect
or
is
g
e
ner
at
e
d
f
or eac
h of t
he
c
ha
nn
e
ls.
3.3.
Opus
Si
m
il
ar
with
Ogg
Vor
bis,
Opus
is
a
f
ull
op
e
n
s
ource
,
non
-
pro
pri
et
ary,
pate
nt
an
d
ro
ya
lt
y
fr
ee
com
pr
essio
n
a
ud
i
o
f
or
m
at
.
It
is
sta
nd
a
rd
iz
e
d
by
the
I
nter
ne
t
Eng
i
neer
i
ng
Task
F
orce
(
I
ETF)
as
R
FC
6716
in
Septem
ber
20
12.
It
is
desi
gned
for
a
wi
de
range
of
a
ppli
cat
ion
s
a
nd
scal
es
from
l
ow
bitrat
e
na
r
rowban
d
sp
eec
h
at
6
kb
it
/s
to
ver
y
high
qual
it
y
ste
reo
m
us
ic
at
51
0
kb
it
/s.
Alm
os
t
si
m
il
ar
with
oth
er
a
ud
i
o
co
de
r,
it
us
es
li
nea
r
pre
dicti
on
an
d
M
DCT.
The
O
pu
s
f
or
m
at
has
th
ree
diff
e
re
nt
m
od
e
s,
i
nclu
ding
s
peec
h,
hybr
id,
a
nd
const
raine
d
en
erg
y
la
pper
t
ra
ns
f
or
m
(CELT
).
T
he
ba
sic
spe
ech
m
od
e
is
usi
ng
S
IL
K
al
gorithm
dev
el
op
ed
by
Sk
ype
m
ai
nly
for
sp
eec
h
sig
nal,
w
hile
CELT
was
us
e
d
m
ai
nly
fo
r
general
aud
i
o
signa
ls.
The
hybri
d
m
od
e
us
es
SI
L
K for t
he
s
peec
h
a
nd
us
es CE
LT
for t
he
f
reque
ncy r
ang
e
abo
ve 8000 Hz
.
3.4.
Free L
os
sle
ss
Au
di
o Codi
ng (FL
AC)
Fr
ee
L
os
sle
ss
aud
i
o
c
od
i
ng
(
FLA
C)
is
on
of
the
m
os
t
popula
r
l
os
sle
ss
c
odec
du
e
it
s
fast
est
dec
od
i
ng
aud
i
o.
FL
AC
use
s
a
li
near
pr
e
dicti
on
(L
P)
operati
on
w
her
e
fu
tu
re
values
of
the
di
gital
si
gn
al
are
est
im
a
te
d
as
a
li
near
f
un
ct
i
on
of
previ
ou
s
sam
ples.
The
FLA
C
e
ncod
er
first
div
ide
the
input
au
di
o
sig
nal
into
f
ram
es.
The
n,
it
will
cond
uct
an
inter
channel
dec
orr
el
at
ion
.
T
he
pr
edict
or
is
util
ized
to
fi
nd
a
n
optim
u
m
coef
fi
ci
ents
to pre
dict t
he
s
ign
al
. L
ast
ly
, th
e
pr
e
dictor
c
oe
ff
ic
ie
nts a
nd i
ts resid
ue were
p
asse
d
t
o
e
ntr
op
y c
odin
g.
3.5.
Tr
ueAu
di
o (T
TA)
Tru
e
A
ud
i
o
is
a
fr
ee,
op
e
n
s
ource
,
an
d
real
tim
e
lossl
ess
aud
i
o
com
pr
ess
or
f
or
m
ulti
ch
ann
el
8,
16,
and
24
bits
au
di
o
data,
with
t
he
abili
ty
of
pas
swor
d
ba
sed
da
ta
protect
ion.
It
was
desi
gn
e
d
to
ha
ve
r
easo
nab
le
com
pr
essio
n
le
vels
wh
il
e
m
ain
ta
inin
g
high
op
e
rati
on
sp
ee
ds
.
T
he
com
pressi
on
rati
o
can
achieve
as
m
uc
h
as
30%
of
ori
gin
a
l fil
e size
, whil
e it
h
as
real t
im
e encodin
g
al
gorithm
.
3.6.
WavP
ack (W
v)
WavPack
is
a
no
t
her
fr
e
e
an
d
op
e
n
s
ource
lossless
a
ud
i
o
c
om
pr
ession
al
gorithm
.
In
the
de
fa
ult
lossless
m
od
e,
W
a
vP
ac
k
act
s
just
li
ke
a
Wi
nZip
c
om
pr
ess
or
f
or
a
udio
file
w
it
h
c
om
pr
ession
rati
o
be
tween
30% to
70%
de
pends o
n
the
a
ud
i
o
s
ource.
T
he hybri
d
m
ode p
r
ovides
a rel
at
ively
s
m
al
l, h
igh
qual
it
y l
os
sy fil
e
that
can
be
use
d
al
l
by
it
sel
f,
an
d
a
co
rr
ect
ion
file
that
pr
ov
i
des
f
ull
los
sle
ss
resto
rati
on.
WavPack
e
m
plo
ys
well
k
no
wn
al
gorithm
s,
su
ch
as
li
near
pre
dicti
on
with
le
ast
-
m
ean
-
sq
ua
red
(LMS)
a
dap
ta
ti
on,
Eli
a
s
an
d
Go
l
om
b
cod
e
s
for
e
ntr
op
y c
odin
g.
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
Perf
orma
nce E
valu
ation of
M
ulti
chann
el
Audio
C
ompre
ssion (
Teddy
Sury
a Gun
awan)
149
4.
RESU
LT
S
AND DI
SCUS
S
ION
This
sect
ion
will
discuss
the
au
dio
data
base
prepa
rati
on,
ex
per
im
ental
set
up
,
i
m
ple
m
entat
ion
,
perform
ance m
et
rics, as
well
as p
e
rfor
m
ance ev
al
uati
on.
4.1.
Experim
en
ta
l
Setu
p
, I
m
plem
ent
at
i
on
and A
udio
Data
ba
se
A
hi
gh
pe
rform
ance
syst
e
m
was
use
d
for
proces
sin
g,
i.e.
a
m
ulti
cor
e
syst
e
m
with
In
te
l
Core
i7
67
00
K
4.00
G
Hz
(
4
c
or
es
wit
h
8
threa
ds),
32
GByt
es
RAM,
256
GByt
es
S
SD
a
nd
2
TBy
te
s
hard
dis
k,
i
ns
ta
ll
ed
with
W
in
dows
10
operati
ng
syst
e
m
and
M
at
la
b
2017b
w
it
h
Sign
al
Pro
cessi
ng
T
oo
l
box.
To
m
ini
m
i
ze
the
eff
ect
of
oth
er
app
li
cat
ions
to
the
sim
ulati
on
,
W
i
ndows
10
was
boote
d
in
Safe
M
ode,
i
n
w
hich
Ma
tl
a
b
was
run
ning
with
no
j
ava
virtu
al
m
achine,
i.e
.
Ma
tl
ab
–
nojv
m
.
Si
m
il
ar
to
[
11
]
,
the
la
te
st
FFm
peg
ve
rsion
3.4.1
was
us
e
d
for
t
he
im
ple
m
enta
ti
on
of
th
ree
lo
ssy
an
d
th
ree
l
os
sle
ss
a
udio
c
od
e
rs.
Ma
tl
ab
syst
e
m
cal
l
dos()
wa
s
us
e
d
to
call
FFm
peg
ex
ecuta
bl
e.
The
au
dio
data
base
was
ext
ra
ct
ed
from
A
m
br
a
E
xp
e
rience
Albu
m
(2
008)
wh
ic
h
ha
s
DT
S
5.
1
(44.1
kH
z
,
16
bits)
a
nd
FLA
C
7.1
(
48
kHz,
16
bits)
f
or
m
at
.
The
ste
reo
sig
nals
wer
e
dow
nm
ix
ed
f
ro
m
the
5.1
au
di
o
so
urce.
O
ut
of
10
trac
ks
,
we
rand
om
l
y
sel
e
ct
ed
three
trac
ks
f
or
ou
r
ex
pe
rim
ents
as
sh
own
in
Ta
ble
2.
No
t
e
that, the
7.1
cha
nn
el
s
h
as
b
i
gger
f
il
e size
due to its
higher
s
a
m
pling
fr
e
que
ncy an
d
ei
gh
t
c
hannels i
n
total
.
Table
2.
A
ud
i
o Data
ba
se fo
r M
ulti
channel
Audio
Au
d
io
Tr
ack
Leng
th
File Size
(
in
MBy
t
es)
2
5
.1
7
.1
Au
d
io
1
0
1
A
m
b
ra
–
Ho
n
o
u
r
an
d
Glory
2
m
in
u
tes an
d
22
seco
n
d
s
2
3
.8
7
1
.5
868
Au
d
io
2
0
3
A
m
b
ra
–
P
ris
m
o
f
L
iv
e
3
m
in
u
tes an
d
28
seco
n
d
s
35
105
152
Au
d
io
3
06
Faszin
atio
n
Natur
–
Sev
en
Seaso
n
s
3
m
in
u
tes an
d
3 seco
n
d
s
3
0
.8
9
2
.5
134
4.2.
Perfo
r
ma
nce
Measures
To
evalu
at
e
the
per
f
or
m
ance
of
au
dio
co
de
rs,
enc
odin
g
ti
m
e
and
per
ce
nt
age
data
redu
ct
ion
wer
e
m
easur
ed
f
or
e
ach
c
oder
an
d
each
a
udio
file
.
F
or
e
ncodin
g
tim
e
(
)
acc
ur
ac
y,
the
Ma
tl
ab
pro
gr
am
will
lo
op
100
ti
m
es
(
)
a
nd
the
a
ve
rag
e
value
is
ta
ke
n
as
s
how
n
in
Eq.
(
1).
T
he
pe
rcen
ta
ge
data
re
du
ct
io
n
is
m
easur
ed
as s
how
n
i
n
E
q. (
2).
(1)
(2)
wh
e
re
is t
he o
rigin
al
file
size i
n byt
es an
d
is t
he
e
nc
od
e
d f
il
e size
in
byte
s.
Fo
r
l
os
sle
ss
a
udio
c
om
pr
essio
n,
s
uch
as
FL
AC,
Tr
ue
Audi
o,
a
nd
Wa
vP
ac
k,
the
re
is
no
loss
in
a
ud
i
o
qu
al
it
y.
But
for
lossy
c
om
pr
e
ssion,
s
uch
as
AA
C,
O
gg
Vorb
is,
an
d
O
pu
s
,
the
re
will
be
loss
in
au
dio
qual
it
y
wh
ic
h
can
be
m
easur
ed
s
ubje
ct
ively
us
ing
li
ste
nin
g
te
st
or
obj
ect
ively
usi
ng
PE
A
Q
[
13]
.
PEAQ
ha
s
been
sta
nd
a
rd
iz
e
d
a
s
IT
U
-
R
BS.
1387
-
1
ha
s
tw
o
m
ai
n
par
ts
,
whic
h
is
t
he
ps
yc
ho
ac
ousti
c
m
od
el
a
nd
the
co
gn
it
iv
e
m
od
el
.
Up
unt
il
now,
PE
AQ
is
only
able
to
m
easur
e
the
obj
ect
iv
e
diff
e
r
ence
gr
a
de
(
O
DG)
for
up
to
ste
reo
sign
al
s.
I
n
[14]
,
the
authors
pro
po
se
d
the
e
xtensi
on
of
PE
AQ
f
or
m
ult
ich
an
nel
au
dio
.
Howe
ver,
it
has
no
t
been
ad
opte
d
a
s
ne
w
sta
nd
a
r
d
ye
t
un
ti
l
no
w.
The
O
DG
sc
ore
ca
n
ra
nge
from
0
to
-
4,
in
wh
ic
h
0
r
ep
res
ents
a
sign
al
with
im
per
ce
ptible
dis
tortio
n
an
d
-
4
r
epr
ese
nts
a
sig
nal
with
ve
ry
annoyi
ng
disto
rtion
.
F
urt
her
m
or
e,
in
this
pa
per
t
he
adv
a
nce
d
ver
si
on
of
PE
AQ
wh
ic
h
ha
s
tw
o
per
i
ph
e
ral
ear
m
od
el
s
an
d
filt
er
bank
base
d
ear
m
od
el
s
was
us
ed
due
to
it
s
ac
cur
acy
.
N
ote
that,
the
curre
nt
PEA
Q
li
m
it
ati
on,
the
PEAQ
m
easur
em
ent
will
be
cond
ucted o
nly o
n
ste
re
o si
gn
al
s and
denote
d
as
as shown
in E
q.
(3).
(3)
wh
e
re is
is t
he
PEAQ
f
un
ct
io
n,
is t
he
origin
al
W
A
V fil
e, a
nd
is t
he
e
ncoded
-
the
n
-
dec
oded
WAV
file
.
4.3.
Ti
me an
d
Fre
quenc
y
S
pectr
um of
Mul
tich
an
nel
Audi
o S
igna
l
Figure
2
s
how
s
the
exam
ple
of
ti
m
e
do
m
ain
an
d
f
reque
nc
y
sp
ect
ru
m
of
7.1
au
dio
sig
na
l
(A
udio
1).
Fr
om
this
fig
ure,
it
can
be
s
een
that
inte
rc
hannel
decorre
la
ti
on
co
uld
be
cond
ucted
be
tween
le
ft
a
nd
ri
gh
t
channel
(
fro
nt,
back,
an
d
si
de
),
i.e.
m
id
an
d
side
si
gn
al
s
,
wh
il
e
f
ront
c
enter
a
nd
LFE
cou
l
d
be
enc
od
e
d
separ
at
el
y as
pract
ic
ed by m
a
ny m
ulti
chan
ne
l aud
i
o
c
oders
.
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.
10
, N
o.
1
,
A
pr
il
2018
:
146
–
153
150
(a)
Fro
nt Lef
t c
hannel
(b)
F
ront Ri
ght
ch
a
nn
el
(c)
Fro
nt Cente
r
c
hannel
(d)
L
ow F
requ
ency cha
nnel
(e)
Bac
k
Le
ft c
hannel
(f) Back
Ri
ght
channel
(g)
Si
de
Le
ft c
hannel
(h)
Si
de
Ri
ght
channel
Figure
2. Sam
ple o
f
tim
e d
oma
in and
fr
e
que
ncy s
pectru
m
of
7.1 a
ud
i
o
sig
nal
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
Perf
orma
nce E
valu
ation of
M
ulti
chann
el
Audio
C
ompre
ssion (
Teddy
Sury
a Gun
awan)
151
4.4.
Experim
en
ts on
L
os
s
y and
Lossless
Com
pression
Table
3
an
d
4
sh
ows
the
da
ta
red
uctio
n
(
%)
an
d
ave
rage
encodin
g
ti
m
e
(seco
nds)
for
lossy
a
nd
lossless
com
pressi
on
of
ste
re
o,
5.1,
an
d
7.1
aud
i
o
sig
nals
us
in
g
A
AC,
O
gg,
O
pus,
FL
AC,
Tr
ue
Audi
o,
a
nd
WavPack
. A
cr
os
s v
ari
ou
s
cha
nn
el
c
onfig
urat
ion
,
the
a
ver
a
ge
da
ta
re
duct
io
n
f
or
lossy
c
om
pr
ession
is 9
1.20%,
92.31%
,
an
d
92.
86%
for
A
A
C,
Ogg
V
orbis
,
and
Opus
.
O
pu
s
has
the
hi
gh
e
st
com
pr
ession
c
om
par
ed
to
the
oth
e
r
al
gorith
m
s.
Me
anw
hil
e,
the
aver
a
ge
data
red
uctio
n
for
lossless
com
pr
ession
is
51
.
63%,
47
.23%,
48.93%
f
or
F
LAC,
Tr
ue
Audio,
a
nd
WavPack,
re
sp
ect
i
vely
.
It
has
be
en
fou
nd
that
FLA
C
has
the
hig
hes
t
com
pr
essio
n
c
om
par
ed
t
o
the
o
the
r
al
gorith
m
s.
Fr
om
Table
3
and
Ta
ble
4,
a
cro
ss
var
i
ou
s
channel
co
nf
i
gurati
ons,
the
a
ver
a
ge
enc
oding
ti
m
e
fo
r
lossy
com
pr
es
sion
is
8.1
9
s
econds
,
5.69
seco
nd
s
,
a
nd
9.46
sec
onds
for
AA
C,
O
gg
Vorb
is
,
a
nd
O
pus
,
resp
ect
ively
.
Me
anwhil
e,
th
e
aver
a
ge
e
ncodin
g
tim
e
fo
r
l
os
sle
ss
c
om
pr
ession
is
1.89
s
econds
,
2.1
6
s
econds
,
and
2
.
11
sec
onds
f
or
FL
AC,
Tru
e
A
ud
i
o,
a
nd W
a
vP
ac
k,
r
e
sp
ect
ively
.
F
or
lossy
com
p
res
sion
o
f
ste
reo
s
ign
al
s,
ano
t
her
p
e
rform
ance
co
uld
be
m
easur
ed
w
hi
ch
is
the q
ua
li
ty
,
i.e.
PE
AQ
O
D
G
.
Ba
se
d
on
Ta
ble
5,
O
gg
Vor
bis
has
t
he hig
hest quali
ty
w
it
h big
m
arg
in c
om
par
ed
to A
AC a
nd Op
us
.
Table
3.
Data
Re
du
ct
io
n (%)
for Vari
ou
s
E
nc
od
e
rs
Enco
d
er
Ch
an
n
el
Au
d
io
1
Au
d
io
2
Au
d
io
3
Av
erage
AAC
2
9
0
.57
9
0
.53
9
0
.8
9
0
.63
5
.1
9
0
.6
9
0
.6
9
0
.72
9
0
.64
7
.1
9
2
.29
9
2
.3
9
2
.41
9
2
.33
Og
g
Vorb
is
2
9
2
.28
9
2
.38
9
2
.75
9
2
.47
5
.1
9
2
.9
9
3
.23
9
4
.19
9
3
.44
7
.1
9
0
.78
9
0
.98
9
1
.33
9
1
.03
Op
u
s
2
9
2
.48
9
2
.8
9
3
.12
9
2
.80
5
.1
9
2
.51
9
3
.01
9
2
.37
9
2
.63
71
9
2
.8
9
3
.32
9
3
.36
9
3
.16
FLAC
2
4
9
.19
4
7
.69
6
0
.58
5
2
.49
5
.1
5
1
.29
5
2
.65
5
8
.49
5
4
.14
7
.1
4
6
.2
4
7
.28
5
1
.34
4
8
.27
Tr
u
eAu
d
io
2
5
0
.17
4
8
.44
6
1
.11
5
3
.24
5
.1
4
1
.93
4
3
.41
5
2
.58
4
5
.97
7
.1
3
9
.01
4
0
.58
4
7
.87
4
2
.49
W
av
Pack
2
4
6
.36
4
4
.81
5
6
.77
4
9
.31
5
.1
4
8
.74
5
0
.82
5
6
.15
5
1
.90
7
.1
4
3
.44
4
5
.18
4
8
.11
4
5
.58
Table
4.
E
nc
od
ing
Tim
e (s
econ
ds)
for Vari
ous E
nc
od
e
rs
Enco
d
ers
Ch
an
n
el
Au
d
io
1
Au
d
io
2
Au
d
io
3
Av
erage
AAC
2
2
.9
4
.39
4
.39
3
.42
5
.1
8
.94
1
2
.62
1
2
.62
9
.82
7
.1
9
.91
1
4
.14
1
4
.14
1
1
.32
Og
g
Vorb
is
2
1
.88
2
.64
2
.64
2
.29
5
.1
5
.33
7
.56
7
.56
6
.39
7
.1
6
.81
9
.85
9
.85
8
.40
Op
u
s
2
1
.29
1
.88
1
.88
1
.76
5
.1
3
.52
5
.05
5
.05
4
.68
71
4
.21
6
.29
6
.29
2
1
.95
FLAC
2
0
.36
0
.48
0
.48
0
.83
5
.1
0
.75
1
.11
1
.11
2
.02
7
.1
1
.11
1
.55
1
.55
2
.83
Tr
u
eAu
d
io
2
0
.42
0
.57
0
.57
0
.89
5
.1
1
.11
1
.55
1
.55
2
.33
7
.1
1
.55
2
.21
2
.21
3
.27
W
av
Pack
2
0
.4
0
.54
0
.54
0
.87
5
.1
0
.99
1
.44
1
.44
2
.24
7
.1
1
.55
2
.11
2
.11
3
.22
Table
5.
O
bj
ec
ti
ve
Q
ualit
y for Vario
us Enc
oders
on St
ere
o Si
gn
al
s
Enco
d
er
Ob
jectiv
e Dif
f
erence Gr
ad
e (
O
DG)
Au
d
io
1
Au
d
io
2
Au
d
io
3
Av
erage
AAC
-
3
.38
-
3
.49
-
3
.38
-
3
.42
Og
g
Vorb
is
-
0
.66
-
0
.76
-
0
.63
-
0
.68
Op
u
s
-
3
.58
-
3
.58
-
3
.56
-
3
.57
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.
10
, N
o.
1
,
A
pr
il
2018
:
146
–
153
152
4.5.
New In
teg
r
ate
d Perf
orm
an
c
e Me
tric
Ba
sed
on
pr
e
vi
ou
s
discuss
i
on,
it
is
rat
her
dif
ficult
to
eval
ua
te
the
perform
ance
of
eac
h
e
ncode
r
as
it
needs
to
eval
ua
te
at
le
ast
two
m
e
tric
s
at
the sam
e
t
i
m
e,
i.e.
encodin
g
ti
m
e
and
data savi
ng,
a
nd q
ualit
y
as
well
in
the
c
ase
of
lossy
com
pr
es
sion
f
or
ste
re
o
sig
nals.
It
is
well
know
n
th
at
there
al
ways
will
be
a
tra
de
-
off
betwee
n
enc
oding
tim
e
(co
m
plexity
)
an
d
da
ta
saving.
Th
e
integrated
m
easur
em
ent
m
e
tric
sh
ould
be
ta
king
care all
of m
easur
em
ent m
et
ri
cs.
We
propose
d
the
foll
owin
g ne
w
inte
gr
at
e
d per
form
ance
m
et
ric:
(4)
wh
e
re
is
the
encodin
g
ti
m
e
(
in
seco
nds)
,
is
the
data
re
du
ct
ion
or
sa
ving
(in
%)
,
an
d
is
the
qual
it
y
(ODG
val
ue)
wh
ic
h
is
only
app
li
cable
f
or
lossy
c
om
pr
ession
(a
nd
up
to
s
te
reo
si
gn
al
s
at
the
m
o
m
ent),
and
is
a
m
easur
em
ent
co
ns
ta
nt
(i
n
seco
nds)
.
Th
e
integrate
d
m
et
ric
was
de
riv
ed
f
ro
m
the
fol
lowing
reas
on
s.
The
perform
ance
of
an
a
udio
e
n
c
od
e
r
is
pr
opor
t
ion
al
to
th
e
da
ta
reducti
on,
,
and
i
nv
e
rsely
pro
portion
al
t
o
and
O
DG
. For t
he
lossless c
om
pr
ession,
t
he
will
b
e d
e
pe
nds
on
ly
on
and
, and
cou
l
d
be
set to a
ver
y
sm
al
l
num
ber
(r
e
pr
ese
nt
ing
a
n
e
xact
r
eplic
a
an
d
no
l
os
s
i
n
in
form
ation).
F
or
our
i
m
ple
m
entat
ion
,
le
t
us
set
for
both
l
os
sle
ss
a
nd
los
sy
com
pr
essio
n,
a
nd
for
los
sle
ss
com
pr
ess
ion
(to
re
pr
ese
nt
hi
gh
qu
al
it
y).
Fr
om
Table
6,
now
we
can
e
valuate
t
he
perform
ance
of
ea
ch
e
nc
od
e
r
i
n
on
e
perform
ance
m
eas
ur
e,
,
pe
r
c
ha
nn
el
co
nf
i
gurati
on
.
T
he
best
pe
rfor
m
ance
bet
ween
cha
nnel
an
d
betwee
n
lossless
a
nd
lossy
com
pr
essio
n
wer
e
hi
gh
li
ght
ed
i
n
bo
l
d.
A
m
on
g
the
loss
y
enc
od
e
rs
for
ste
re
o
sig
nals
,
O
gg
V
orbis
has
t
he
highest
pe
rfo
r
m
ance.
H
owe
ver,
O
pus
has
the
hi
ghest
pe
rfor
m
ance
f
or
5.1
an
d
7.1
cha
nn
el
.
Am
ong
t
he
lossless
com
pressi
on,
FLA
C
has
the
highes
t
per
f
or
m
ance
for
al
l
chan
nel
con
fi
gurati
on.
In
co
nclusi
on,
our
integrate
d
m
easur
e
,
,
is
a
ble
to
ca
pture
t
he
perform
ance
of
eac
h
enc
oder
in
te
rm
s
of
e
ncodin
g
tim
e,
d
at
a
savin
g, an
d q
ua
li
ty
.
Table
6.
Per
for
m
ance Ev
al
uation o
f Vario
us
Enc
od
e
rs usin
g I
ntegr
at
e
d
Me
tric
Enco
d
er
Files
Ch
an
n
el
2
5
.1
7
.1
AAC
Au
d
io
1
9
.2
1
0
.1
9
.3
Au
d
io
2
5
.9
7
.2
6
.5
Au
d
io
3
6
.1
7
.2
6
.5
Averag
e
7
.1
8
.2
7
.5
Og
g
Vorb
is
Au
d
io
1
7
4
.4
1
7
.4
1
3
.3
Au
d
io
2
4
6
.0
1
2
.3
9
.2
Au
d
io
3
5
5
.8
1
2
.5
9
.3
Averag
e
5
8
.7
1
4
.1
1
0
.6
Op
u
s
Au
d
io
1
2
0
.0
2
6
.3
2
2
.0
Au
d
io
2
1
3
.8
1
8
.4
1
4
.8
Au
d
io
3
1
3
.9
1
8
.3
1
4
.8
Averag
e
1
5
.9
2
1
.0
1
7
.2
FLAC
Au
d
io
1
1
3
6
6
.4
6
8
3
.9
4
1
6
.2
Au
d
io
2
9
9
3
.5
4
7
4
.3
3
0
5
.0
Au
d
io
3
1
2
6
2
.1
5
2
6
.9
3
3
1
.2
Averag
e
1
2
0
7
.3
5
6
1
.7
3
5
0
.8
Tr
u
eAu
d
io
Au
d
io
1
1
1
9
4
.5
3
7
7
.7
2
5
1
.7
Au
d
io
2
8
4
9
.8
2
8
0
.1
1
8
3
.6
Au
d
io
3
1
0
7
2
.1
3
3
9
.2
2
1
6
.6
Averag
e
1
0
3
8
.8
3
3
2
.3
2
1
7
.3
W
av
Pack
Au
d
io
1
1
1
5
9
.0
4
9
2
.3
2
8
0
.3
Au
d
io
2
8
2
9
.8
3
5
2
.9
2
1
4
.1
Au
d
io
3
1
0
5
1
.3
3
8
9
.9
2
2
8
.0
Averag
e
1
0
1
3
.4
4
1
1
.7
2
4
0
.8
5.
CONCL
US
I
O
NS
A
ND FUT
UR
E
WO
RKS
This
pa
pe
r
has
pr
ese
nted
t
he
perform
ance
evaluati
on
of
th
ree
lossy
an
d
t
hr
ee
l
os
sle
ss
c
om
pr
essio
n
for
m
ult
ic
hann
el
aud
io
sig
nal
s,
inclu
ding
ste
reo,
5.1
an
d
7.
1
cha
nn
el
s
.
Th
e
six
aud
i
o
co
m
pr
ession
al
go
rithm
,
i.e.
A
AC,
O
gg
V
orbis,
O
pu
s
,
FLA
C
,
T
ru
e
A
ud
i
o,
a
nd
WavPack,
ha
ve
bee
n
c
onfirm
ed
to
be
a
ble
to
pe
r
form
com
pr
essio
n
up
to
7.1
cha
nn
el
.
Ex
per
im
ents
we
re
c
onduct
ed
on
the
s
am
e
three
au
dio
file
s
but
wit
h
di
f
fer
e
nt
channel
c
onfi
gurati
ons.
T
he
perf
or
m
anc
e
of
each
enc
od
e
r
was
e
valuated
base
d
on
it
s
enc
odin
g
ti
m
e
(av
e
rag
e
d
ove
r
10
0
ti
m
es),
data
sa
ving,
an
d
a
ud
i
o
qual
it
y.
Further
m
or
e,
we
pro
po
s
ed
one
i
ntegr
at
e
d
perform
ance
m
easur
e
to
e
ase
the
e
valua
ti
on
.
Using
th
e
new
m
easure
,
F
LAC
was
fou
nd
to
be
th
e
bes
t
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
Perf
orma
nce E
valu
ation of
M
ulti
chann
el
Audio
C
ompre
ssion (
Teddy
Sury
a Gun
awan)
153
lossless
c
om
pr
ession,
wh
il
e
Ogg
V
orbis
a
nd
Opus
we
re
f
ound
to
be
the
best
for
l
os
sy
com
pr
essio
n.
Fu
rt
he
r
researc
h
ca
n
be
cond
ucted
i
nc
lud
in
g
the
use
of
dif
fer
e
nt
aud
i
o
data
base
,
pa
ram
et
ers
op
tim
iz
at
ion
for
each
encode
r,
a
nd th
e u
se
of
diff
e
re
nt audio
cod
e
rs
.
ACKN
OWLE
DGE
MENT
The
aut
hors
w
ou
l
d
li
ke
to
express
their
grat
it
ud
e
to
the
Ma
la
ysi
an
Min
ist
ry
of
High
er
Ed
ucati
on
(MO
HE),
w
hich
has
pro
vid
e
d
f
unding
f
or
the
researc
h
th
rou
gh
the
F
un
dam
ental
Re
se
arch
Gr
a
nt
Sc
hem
e,
FRGS
15
-
194
-
0435.
REFERE
NCE
S
[1]
M.
Bosi
and
R.
E.
Goldber
g
,
“
Introduc
ti
on
to
d
igi
tal
audi
o
coding
and
standa
rd
s,
”
Springer
Sci
enc
e
&
Business
Media
,
vol
.
721
,
2012.
[2]
F.
Rum
se
y
,
“
Spati
al a
ud
io,
”
CRC
Press
,
2012.
[3]
F.
Schuh,
et
al
.
,
“
Eff
ic
i
ent
m
ult
ic
h
annel
aud
io
tra
nsform
coding
with
low
de
lay
and
c
om
pl
e
xity
,
”
in
Audi
o
Engi
ne
ering
Soc
ie
t
y
Conv
ent
ion
,
pp.
141
,
2016
.
[4]
J.
Moffitt,
“
Ogg
Vorbis
—
open,
f
re
e audi
o
—
se
t your media free,
”
Linux
journal
,
v
ol.
2001
,
pp
.
9
,
2
001.
[5]
W
.
Jac
kson,
“
Audio
Con
ce
p
ts,
Te
rm
inol
og
y
,
an
d
Codec
s,
”
in
A
ndroid
Apps
for
Abs
olut
e
Begi
nn
ers:
Springer,
pp
.
651
-
663
,
2014
.
[6]
M
.
Bosi,
et
al.
,
“
ISO
/IE
C
MP
E
G
-
2
adva
nc
ed
a
udio
cod
ing,
”
Jo
urnal
of
the
Aud
io
eng
ine
ering
society
,
vol
.
45
,
p
p
.
789
-
814,
1997
.
[7]
R.
Henne
qu
in,
e
t
al.
,
“
Codec
in
depe
nden
t
loss
y
audi
o
compre
ss
ion
detec
ti
on,
”
i
n
Ac
ousti
cs,
Sp
e
ec
h
and
S
ignal
Proce
ss
ing
(
ICASSP
)
,
2017
IE
E
E
Int
ernati
onal
Confe
renc
e
on
,
pp.
726
-
730
,
20
17.
[8]
T.
Liebc
hen
,
et
al.
,
“
The
MP
EG
-
4
Audio
Lossless
Coding
(
ALS
)
standa
rd
-
techn
olog
y
and
app
l
icati
ons,
”
in
Proc.
119th
A
ES
Conv
,
20
05
.
[9]
A.
Djuric,
“
T
TA
Lossless a
udio
c
odec
-
Tr
u
e audi
o
compress
or
al
g
orit
hm
s,
”
2010
.
[10]
T.
S.
Gunawan,
et
al
.
,
“
Inve
stigation
of
Lossless Audio
Com
pre
ssi
on
using I
E
EE
1857.
2
A
dvance
d
Audio
Coding
,
”
Indone
sian J
our
nal
of
Elec
tric
al
Engi
ne
ering
and
Computer
Sc
ie
n
c
e
,
vol
.
6
,
2017
.
[11]
T
.
S.
Gunawan,
et
al.
,
“
Inve
stigation
of
Vari
ous
Algorit
hm
s
on
Multi
cha
nn
el
A
udio
Com
pre
ss
ion,
”
in
Proc
.
of
the
4th
IEE
E
Int
ernati
onal
Confe
ren
ce
on
Smar
t
Instrum
ent
ati
on,
Me
asur
eme
nt
and
Appl
ic
at
ions
(
ICS
IMA
)
,
Putraja
y
a
,
pp.
1
-
5
,
2017
.
[12]
T.
Holm
an
,
“
Surround sound:
up
and
running
,
”
C
RC Pre
ss
,
2014.
[13]
T.
S.
Gunawan,
“
Audio
compres
sion
and
spee
ch
enha
nc
ement
using
te
m
pora
l
m
asking
m
odel
s,
”
T
hesis,
Univer
sit
y
of
New South
W
al
es,
S
y
dn
e
y
,
Au
stral
i
a, 2007.
[14]
J.
Li
ebetra
u
,
et
a
l.
,
“
Standa
rd
ization
of
PEAQ
-
M
C:
Ext
ension
of
ITU
-
R
BS
.
1387
-
1
to
m
ult
ic
hann
el
audi
o
,
”
in
A
E
S
40th
Int
.
Con
f, S
pati
al Audi
o:
Se
nse
the Sound
o
f
Space
,
Tok
y
o,
Japa
n,
2010.
Evaluation Warning : The document was created with Spire.PDF for Python.