Int
ern
at
i
onal
Journ
al of Ele
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
8
,
No.
6
,
D
ece
m
ber
201
8
, pp.
4356
~
43
65
IS
S
N: 20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v8
i
6
.
pp
4356
-
43
65
4356
Journ
al h
om
e
page
:
http:
//
ia
es
core
.c
om/
journa
ls
/i
ndex.
ph
p/IJECE
Linea
r Ph
ase FIR L
ow
P
ass Filte
r Design
Based on
Firefly
Algorith
m
Moa
th
S
abab
ha
,
Moh
ame
d
Z
oh
dy
Elec
tr
ical and
C
om
pute
r
Engi
n
e
eri
ng,
Oakl
and Univ
ersity
,
US
A
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
N
ov
16
, 201
7
Re
vised
Ju
l
3
0
,
201
8
Accepte
d
Aug 13
, 201
8
In
thi
s
p
ape
r
,
a
li
ne
ar
phase
Lo
w
Pass
FIR
fil
ter
is
design
ed
an
d
propo
sed
base
d
on
Firef
l
y
al
gori
thm.
W
e
expl
o
it
th
e
e
xploi
tation
and
expl
ora
ti
on
m
ec
hani
sm
with
a
lo
cal
sea
rch
r
outi
ne
to
impro
ve
th
e
conv
erg
e
nce
and
ge
t
highe
r
spe
ed
co
m
puta
ti
on.
Th
e
opti
m
um
FIR
fil
te
rs
are
designed
base
d
o
n
the
Firefly
m
ethod
for
which
the
fini
t
e
word
le
ngth
is
used
to
rep
rese
nt
coe
ffi
ci
en
ts.
Furthermore,
Pa
rti
cle
Sw
arm
Opti
m
iz
at
ion
(PSO)
and
Diffe
ren
t
ia
l
Evo
lut
ion
al
gor
it
hm
(DE)
will
be
use
d
to
show
th
e
so
lut
ion.
Th
e
result
s
will
b
e
c
om
par
ed
with
PS
O
and
DE
m
et
hods.
Firefly
a
lg
orit
hm
and
Parks
–
McCle
lla
n
(PM
)
al
gorit
hm
are
al
so c
om
par
ed
in
th
is pa
pe
r
thoroughly
.
The
design
go
al
is
succ
essfull
y
ac
hi
eve
d
in
all
design
exa
m
pl
es
using
the
Firefly
al
gori
th
m
.
They
ar
e
co
m
par
ed
with
tha
t
obta
ine
d
b
y
us
ing
the
PS
O
and
the
DE
a
lgo
rit
hm
.
For
the
p
roble
m
at
hand
,
the
sim
ula
ti
on
r
esult
s
show
tha
t
th
e
Firefly
al
gorit
hm
outper
form
s
the
PSO
and
DE
m
et
hods
in
som
e
of
the
pre
sen
te
d
d
esign
exa
m
ple
s.
It
al
so
per
for
m
s
well
in
a
porti
on
of
t
h
e
exhi
bited
d
esign
exa
m
ple
s
par
t
icular
l
y
in
sp
ee
d
a
nd
qualit
y
.
Ke
yw
or
d:
Conver
ge
nce
Diff
e
re
ntial
Ev
olu
ti
on
(D
E
)
Finit
e Im
pu
lse
Fil
te
r
(FIR)
Firefly
alg
or
it
hm
Lo
w pass
filt
er
Par
ks
–
Mc
Cl
el
la
n (PM)
Partic
le
Sw
a
rm
Optim
iz
at
ion
(P
S
O)
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
:
Moat
h
Saba
bh
a,
Ele
ct
rical
an
d
Com
pu
te
r
E
ng
i
neer
i
ng,
Oak
la
nd Unive
rsity
, Ro
c
hester,
MI
48
309 U
SA
.
Em
a
il
:
m
sabab
ha@oa
klan
d.
e
du
1.
INTROD
U
CTION
Durin
g
process
ing
of
sig
nal,
unwa
nted
pa
rts
of
the
si
gn
al
ar
e
rem
ov
ed
by
us
e
of
a
filt
er.
This
pape
r
looks
at
the
di
gital
filt
er.
It
use
s
dig
it
al
proc
essors
to
car
ry
ou
t
al
gorit
hm
on
dif
fer
e
nt
sa
m
ples
of
the
va
lue
of
the
sig
nal.
Th
e
dig
it
al
filt
er
br
in
gs
a
rithm
et
ic
al
pr
oce
dures
on
m
od
el
s
of
un
m
ist
akab
le
tim
e
sign
al
s
t
o
dim
inish
or
e
nhance
highli
gh
ts
of
the
sig
nals.
Th
e
get
ri
d
of
nu
m
ero
us
pro
blem
s
associ
at
ed
with
ad
diti
on
a
l
pr
i
nciple
cl
ass
of
el
ect
r
on
ic
F
il
te
rs
su
ch
as
the
A
nalo
gu
e
t
o
Digital
Conv
erter
(
AD
C
).
T
her
e
are
t
wo
cl
asses
of
dig
it
al
sign
a
ls. Th
at
is, the
finite
i
m
pu
lse
re
spon
se
on
(FI
R) an
d
t
he
infi
nite im
pu
ls
e resp
onse
(IIR) si
gn
al
s
.
They
are
ide
ntifie
d
with
the
l
eng
t
h
of
the
im
pu
lse
resp
on
se.
The
F
IR
sign
al
is
an
al
lur
ing
ch
oice
bec
ause
of
it
s
effor
tl
essne
ss
in
desig
n
a
nd
ste
adi
ness.
By
plo
tt
ing
th
e
signa
l
val
ve
s
to
be
e
qu
i
va
le
nt
with
t
he
m
idd
le
valve
area
,
t
he
FI
R
filt
ers
ha
ve
a
li
near
pha
se.
FI
R
cha
nn
el
s
are
recogn
i
zed
to
co
ntain
nu
m
erous
ap
pe
al
ing
at
tribu
te
s
li
ke
certai
n
c
onsist
ency,
t
he
c
hance
of
e
xact
li
ne
ar
ph
a
se
featur
e
at
al
l
f
re
quencies
an
d
ad
van
c
e
d
execu
ti
on
as
non
-
rec
ursive
de
velo
pm
ents [
1
]
-
[
9].
Var
i
ou
s
str
at
egies
are
em
plo
ye
d
for
the
ar
rangem
ent
of
dig
it
al
filt
ers
[
10
]
,
[
11]
.
The
windowi
ng
strat
egy
is
the
m
os
t
pr
efer
red.
I
n
this
strat
eg
y,
the
pe
rf
ect
i
m
pu
lse
response
is
increase
d
by
bein
g
m
ult
ipli
ed
with
a
window’s
f
un
ct
io
n.
Ther
e
are
va
riou
s
ass
or
tm
ents
of
ty
pe
s
of
wind
ow
ca
pa
ci
ti
es
(Butterworth
,
Chebys
hev,
K
ai
ser
an
d
s
o
on.
).
T
he
ty
pes
dep
e
nd
on
t
he
requirem
ents
of
ri
pp
le
s
on
th
e
pass
ba
nd
an
d
sto
p
band.
Als
o,
t
he
y
dep
e
nd
on
the
sto
p
ba
nd
reducti
on
a
nd
the
co
nversi
on
widt
h.
These
disti
nctive
win
dows
bound
the
i
nf
i
nite
le
ng
t
h
dr
i
ve
respo
ns
e
of
ideal
filt
ers
i
nt
o
a
finite
wi
ndow
t
o
desig
n
a
ge
nu
i
ne
respon
s
e.
I
n
any
case,
wi
ndowin
g
syst
em
s
do
n'
t
al
low
e
nough
co
m
po
sit
ion
of
t
he
fr
e
qu
e
ncy
respo
ns
e
i
n
th
e
div
e
rs
e
fr
e
qu
e
ncy
bands
a
nd
ot
her
filt
er
facto
r
s
s
uch
as
the
pr
ogress
widt
h.
The
Re
m
ez
-
exch
a
nge
al
go
rithm
s
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Linear P
hase
FIR Low
Pass
Fil
te
r D
esi
gn
Base
d on
..
.
(
M
oa
t
h Sab
abha
)
4357
(trad
it
io
nal
te
chn
i
qu
e
s)
int
rodu
ce
d
by
Par
ks
an
d
Mc
Cl
el
la
n
PM
do
e
s
no
t
per
m
it
e
xpress
c
ho
ic
e
of
th
e
fr
e
qu
e
ncy
re
spon
s
e
co
ntaine
d
in
the
disti
nc
t
fr
e
qu
e
ncy
ba
nd
s
.
Tec
hniq
ue
s
base
d
on
t
he
Rem
ez
exch
ang
e
al
gorithm
are
t
he
m
os
t
pr
i
m
itive
ones
[
12
]
.
Howe
ver,
this
al
gorithm
do
es
no
t
gi
ve
r
oo
m
fo
r
e
xpli
ci
t
selecti
on
of
the
t
op
li
m
i
t
of
the
abs
olut
e
rip
ple
in
the
pass
ba
nd
a
nd
sto
pb
a
nd
(δp,
δs
),
rathe
r
on
e
can
just
dete
rm
ine
their
pro
portio
n.
M
or
e
ov
e
r,
t
he
PM
giv
es
dri
fting
point
c
oe
ff
ic
ie
nts
wh
ic
h
re
quires
qu
a
ntiza
ti
on
if
e
qu
ipm
ent
execu
ti
on
is
l
ooke
d
f
or.
Ba
sed
on
these
r
easo
ns
,
w
hen
desig
ning
li
ne
ar
phase
FI
R
filt
ers,
s
om
e
dig
it
al
stochastic
wor
ldwid
e
m
od
el
al
gorithm
su
ch
as
the
Diff
e
r
entia
l
Ev
olu
ti
on
(
DE
)
al
gorit
hm
,
Partic
le
Sw
arm
Op
ti
m
iz
ation
(PSO
),
a
nd
Genet
ic
A
lgorit
hm
(
G
A) are
u
se
d [13]
-
[
15]
.
The
sig
nifica
nt
lim
it
ation
of
the
desi
gn
i
ng
proce
dure
is
tha
t
the
si
m
i
la
r
est
i
m
a
ti
on
s
of
th
e
a
m
plit
ud
e
m
ist
ake
in
the
fr
e
qu
e
ncy
ba
nds
a
re
s
pecific
by
m
eans
of
t
h
e
weig
htin
g
c
apacit
y,
an
d
not
by
the
div
e
r
gen
ce
s
them
sel
ves.
Along
these
li
nes
,
if
there
s
houl
d
arise
an
occ
urre
nce
of
sche
m
ing
ba
nd
filt
ers
with
a
kn
own
sto
p
band
uniq
ue
ne
ss,
filt
er
le
ngth
an
d
c
ut
-
off
re
currence
,
t
he
a
rr
a
ng
em
ent
m
us
t
be
it
erate
d
sever
al
ly
.
N
um
ero
us
cop
ie
s
ha
ve
be
en
e
xten
de
d
for
the
(
FI
R)
sig
nal
te
ch
niques
and
pla
n
desi
gns.
This
is
a
n
i
nv
e
sti
gative
a
r
ea
that
aim
s
at
acco
m
plishin
g
m
or
e
gen
e
ral
an
d
s
pe
arh
ea
ding
str
at
egies
that
ar
e
com
petent
to
determ
ine
as
well
as
adv
a
nce
ne
w
a
nd co
m
pound des
ig
ning tec
hniq
ues [16]
.
Diff
e
re
nt
co
nventio
nal
te
c
hniq
ues
e
xist
f
or
di
gital
FI
R
cha
nn
el
desi
gn.
Ou
t
of
th
os
e
one
is
the
de
sig
ning
of
the
filt
er
usi
ng
t
he
dig
it
al
al
gorithm
s.
T
his
te
ch
nique
has
a
stu
nnin
g
capaci
ty
.
T
ha
t
is,
it
giv
es
al
te
r
natives
of
util
iz
a
tio
n
of
va
rio
us
dig
it
al
al
gorith
m
s
and
ind
ece
ncy
in
c
onfig
urat
ion.
It
s
peci
fical
ly
reli
es
upon
ex
ecuti
on
of
al
gorithm
s.
At
fir
st
this
te
chn
iq
ue
was
us
e
d
by
Park
s
a
nd
Mc
Cl
el
la
n
by
util
iz
ing
a straig
htf
orwa
rd
it
erati
ve
s
uburba
nite pro
gr
a
m
an
d
is nam
e
d
as
PM
m
et
ho
d
f
or f
il
te
r desi
gn
i
ng. T
his str
at
egy
was
al
te
red
la
te
r
on
by
sup
planting
t
he
us
e o
f
basic
pro
gr
a
m
with
enh
a
nc
e
m
ent
al
go
rith
m
s.
At
first
her
edita
r
y
al
gorithm
was
util
iz
ed
f
or
a
wide
ra
nge
of
c
ha
nn
e
l
desig
n.
T
hi
s
was
tr
ai
le
d
by
util
iz
at
ion
of
diff
e
re
n
t
al
gor
it
h
m
s.
As
of
la
te
,
cro
ssov
er
al
gorithm
and
a
dju
ste
d
al
gorithm
s
has
been
c
reated
from
fun
dam
ental
s
or
t
of
al
go
rith
m
fo
r
c
hanges
in
the
sig
nal
de
sign
s
.
Em
plo
ym
ents
of
en
ha
nced
m
olecul
e
swar
m
adv
a
ncem
ent,
ver
sat
il
e
dev
el
op
m
ent
m
olec
ule
s
war
m
i
m
pr
ovem
en
ts
for
the
filt
er
desi
gnin
g
validat
e
t
he
la
te
r
tren
ds
[16
]
-
[
24].
Seve
ral
ap
pro
aches
wer
e
pr
esented
to
de
sign
li
near
ph
ase
FI
R
filt
er
s.
F
or
insta
nc
e,
sim
ulate
d
ann
eal
in
g
a
nd
GA
m
et
ho
d
w
ere
a
pp
li
ed
to
desig
n
F
IR
filt
ers
with
c
oef
fi
ci
ents
val
ues
e
xpresse
d
as
a
p
ow
e
r
of
two.
H
ow
e
ve
r
,
these
a
ppr
oa
ches
a
re
co
m
pu
ta
ti
on
al
ly
ver
y
e
xp
e
ns
i
ve
.
The
Firefly
m
et
ho
d
is
e
asy
to
i
m
ple
m
ent,
an
d
it
s
c
onverge
nce
is
c
ontroll
ed
via
fe
w
para
m
et
ers.
H
ow
e
ver,
o
ther
m
eth
ods,
s
uch
as
Gen
et
ic
and
D
E
al
gorit
hm
s,
involves
search
pr
ocedu
res
us
in
g
t
he
popula
ti
on
ge
ne
ti
cs
and
nat
ur
a
l
sel
ect
ion
pro
cess.
The
pur
po
se
of
this
pa
per
is
to
us
e
the
Fire
fly
appr
oach
a
s
an
al
te
r
nativ
e
m
e
tho
d
to
t
he
se
ap
proac
hes
due
to
it
s
ease
of
im
ple
m
entat
ion
in b
ot
h
par
am
et
er
sel
ect
ion
a
nd
t
he
c
on
te
xt
of
c
od
i
ng.
T
his
pa
per
prop
os
es
a li
near
ph
a
se
Lo
w
Pas
s
FI
R
filt
er
ba
s
ed
ona
n
al
te
rnat
ive
appr
oach
cal
le
d
Firefly
m
et
ho
d
that
as
su
res
t
he
r
obust
nes
s
of the
FI
R
desi
gn in
te
rm
s o
f per
form
ance an
d com
pu
ta
ti
onal
co
m
plexity
.
This
pa
per
is
orga
nized
asf
ol
lows
.
I
n
Sect
i
on
II
,
t
he
syst
e
m
design
an
d
analy
sis
of
th
epro
po
se
d
li
near
phase
F
IR
low
pass
fi
lt
er
are
pr
e
sen
te
d.
The
Dif
fe
ren
ti
al
Ev
olu
ti
on
(DE)
al
gori
thm
and
the
Partic
le
Sw
arm
al
go
rit
hm
(P
SO
)
a
re
discusse
d
br
ie
f
ly
in
Sect
ion
I
I
I
an
d
Sect
io
n
I
V
,
resp
ect
i
vely
.
I
n
Sect
io
n
V
,
Th
e
Firefly
al
gorith
m
is
pr
esente
d.
Si
m
ulati
on
an
d
E
xam
ples
of
Desig
n
F
IR
usi
ng
Firefly
al
gorithm
pr
esen
te
d
i
n
Sect
ion
V
I.
Fin
al
ly
, co
ncl
us
io
ns
a
re
ou
tl
ine
d i
n
Sect
io
n VII.
2.
THE
WIN
D
OWS
F
RAM
EWOR
K
O
U
TL
INE
A
ND
E
X
A
MI
NA
T
ION
O
F
TH
E
PROP
OSE
D
LINEA
R
P
H
AS
E FI
R
W
i
nd
owin
g
str
at
egy
is
us
ed
to
get
the
sim
p
le
ty
pe
of
FI
R
filt
ers.
FI
R
co
nf
i
gurati
on
be
gin
s
with
a
perfect
w
a
nte
d fr
e
quency
res
pons
e
obtai
ne
d as f
ollo
ws:
(
)
=
∑
ℎ
[
]
−
∞
=
−
∞
(1)
Wh
e
re;
ℎ
[
]
is
the
i
m
pu
lse
res
pons
e
of
th
e
disti
nct
cha
nnel
.
T
hen,
we
nee
d
t
o
m
ake
ℎ
[
]
a
cau
sal
FI
R
Fil
te
rb
y
two
sta
ges.
The
first
ℎ
[
]
is
increase
d
by
a
finite
le
ngth
a
gr
eem
ent
[
]
(w
i
ndow)
ke
epin
g
i
n
m
ind
the
en
d
goal
of
getti
ng
a
finite
le
ng
t
h
im
pu
lse
re
sp
on
se.
The
sec
ond
on
e
wh
ic
h
is
c
ausali
ty
is
pr
es
ented
by tim
e d
efe
rr
i
ng the
wind
owed dri
ve
re
spo
ns
e.
One ca
n
e
xpress
t
his in
m
at
he
m
at
ic
a
l ter
m
s as f
ollo
w
s
:
ℎ
[
]
=
ℎ
[
]
[
]
(2)
Wh
e
re
ℎ
[
]
denotes
the
wind
ow
e
d
i
m
pu
lse
res
pons
e;
so,
the
f
re
qu
e
ncy
res
ponse
of
t
he
wi
ndow
e
d
im
pu
lse
respo
ns
e
(
)
is
the
per
i
od
ic
c
onvo
l
ution
of
the
desire
d
f
re
quency
res
pons
e
(
)
with
the
Fouri
e
r
trans
form
Γ
(
)
of th
e w
in
dow
and i
s g
i
ven b
y
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
87
08
In
t J
Elec
&
C
om
p
En
g,
V
ol.
8
, N
o.
6
,
Dece
m
ber
2
01
8
:
4356
-
4365
4358
()
1
(
)
(
)
(
)
.
2
iw
i
i
w
wd
H
e
H
e
e
d
(3)
The fre
quency
respo
ns
e
of the
ap
pro
xim
a
ti
on
f
il
te
r
will
be
(
)
(
)
d
iwm
iw
iw
w
H
e
e
H
e
(4)
Wh
e
re
d
m
re
pr
ese
nts
the
necess
ary
tim
e
dela
y
to
intro
duce
causali
ty
of
t
he
ap
prox
im
ated
filt
er.
Fil
te
rs
desig
ne
d
by
w
indowi
ng
will
hav
e
great
est
error
on
ei
ther
side
of
the
di
sco
ntinu
it
y
of
the
ideal
f
requ
ency
respo
ns
e
a
nd
s
m
al
le
r
err
or
for
f
re
qu
e
ncies
a
way
f
ro
m
the
disco
ntin
uity
.
The
fr
e
quency
res
pons
e
()
iw
He
of
an
M l
e
ng
t
h
fi
nite w
ord
-
le
ng
t
h
F
IR li
nea
r pha
se d
i
gital
f
il
te
r
is
giv
e
n
as:
1
0
(
)
[
]
M
iw
iw
k
d
k
H
e
h
k
e
(5)
Wh
e
re
[
]
.
.
.
0
,
1
,
2
,
.
.
.
.
,
1
h
k
k
M
are
b
-
bi
t
(sign
bit
inc
lud
e
d)
filt
er
c
oeffici
ents.
P
r
efera
ble
cha
nn
el
s
ov
e
r
the
windowin
g
strat
egy
r
esult from
the
m
ini
m
iz
at
ion
of m
axi
m
u
m
err
or
yi
el
ds
the m
os
t p
ref
e
rab
le
f
il
te
rs
ov
e
r
the
wind
ow
i
ng
m
et
ho
d.
These
kinds
of
filt
ers
can
be
go
tt
en
by
util
iz
ing
al
go
rithm
ic
procedure
s.
In
a
n
effor
t
t
o
ou
tl
in
e
FI
R
filt
ers
i
n
wh
ic
h
s
om
e
of
t
he
par
am
et
ers,
desi
gn
e
d
a
lgorit
hm
s
are
dev
el
op
e
d.
S
om
e
of
these
par
am
et
e
rs
incl
ud
e;
the
channel
le
ng
t
h
(
),
pass
band
a
nd
sto
pban
d
st
and
a
r
dized
fr
e
qu
e
ncies
(
,
).
Re
searche
rs
ha
ve
de
velo
pe
d
al
gorithm
s
in
wh
ic
h
,
,
an
d
are
fixe
d.
T
he
oth
e
r
param
et
e
rs
ha
ve
bee
n
enh
a
nce
d.
Dif
f
eren
t
est
im
at
es
wer
e
i
niti
al
ly
set
up
by
Pa
r
ks
a
nd
Mc
Cl
el
la
n
in
w
hic
h
,
,
,
an
d
t
he
pro
portion
δ
s
⁄
is
f
ixed.
From
that
po
int
forw
a
r
d,
the
Pa
rk
s
–
M
cC
le
ll
an
(P
M)
al
gorithm
is
the
m
os
t
fam
ou
s
appr
oach
f
or
i
deal
FI
R
filt
ers
desig
n.
T
his
is
so
because
of
it
s
a
dap
ta
bili
ty
and
com
puta
ti
on
al
pro
duc
ti
vity
.
An app
roxim
ate
erro
r
f
unct
i
on in
the
PM al
gorithm
is charac
te
rized
by:
(
)
(
)
(
)
(
)
iw
iw
d
E
w
G
w
H
e
H
e
(6)
Wh
e
re
()
iw
d
He
is
the
fr
e
quency
re
sp
onse
of
the
desire
d
filt
er
an
d
()
iw
He
ist
he
ap
pro
xim
a
te
filter,
resp
ect
ively
,
(
)
is
a
weig
hing
functi
on.
It
is
us
e
d
to
giv
e
the
wei
gh
ti
ng
of
t
he
ap
pro
xim
a
ti
on
er
ror
disti
nctivel
y
in
var
io
us
rec
urr
ence
gro
ups
of
fr
eq
ue
ncy
ba
nd
s
.
The
obj
ec
ti
ve
of
the
de
s
ign
is
to
locat
e
the
appr
ox
im
at
e
fi
lt
er
coeffic
ie
nts
that
are
outc
om
es
of
the
id
eal
filt
er.
Sinc
e
t
he
m
axi
m
u
m
err
or
is
m
ini
m
iz
ed,
then
the
filt
er
is
op
ti
m
u
m
.
In
m
at
he
m
atical
te
r
m
s,
the
best
app
r
oxim
ati
on
is
to
be
found
in
the
f
ollow
i
ng
sense:
|)
)
(
|
m
a
x
(
m
i
n
}
0
:
]
[
w
E
F
w
M
n
n
h
(7)
Wh
e
re
)
(
n
h
is
the
i
m
pu
lse
res
ponse
of
the
a
ppr
oxim
a
te
filt
er
a
nd
is
the
cl
os
e
d
subset
of
0
≤
≤
0
.
5
.
This
m
od
el
do
es
not
pe
rm
it
e
xpress
determ
i
nation
of
t
he
m
os
t
ext
rem
e
of
the
outri
gh
t
rip
ples
in
t
he
pass
ba
nd
and
sto
pban
d
(
,
),
rathe
r
on
e
can
just
in
dica
te
their
pro
por
ti
on
.
Mo
reover
,
the
PM
giv
e
s
dri
ft
in
g
poin
t
coeffic
ie
nts
w
hich
re
quire
quantiz
at
io
n
if
hard
war
e
im
pl
e
m
entat
ion
is
so
ug
ht.
This
m
ot
ivate
s
util
i
zi
ng
a
recently
de
vel
op
e
d
stoc
hastic
gl
ob
al
opti
m
iz
at
ion
al
go
rithm
cal
le
d
Fire
fly
al
gorithm
to
desig
n
li
near
phase
FI
R
filt
er
in
t
his
pa
per
. Th
e Fir
efly
is
an
inse
ct
that
m
os
tl
y
pro
du
ces short an
d
r
hythm
ic
f
la
sh
es
that
pr
oduc
e
d
by
a
process
of
bio
l
um
inescence.
T
he
f
unct
ion
of
t
he
flas
hi
ng
li
ght
is
to
at
tract
par
tne
rs
(co
m
m
un
ic
at
i
on)
or
at
tract
p
otentia
l pr
ey
an
d
as a
protect
ive w
a
r
ning tow
a
r
d
th
e p
re
d
at
or. Thus,
this inte
ns
it
y of
li
gh
t i
s the f
act
or
of the
oth
e
r fir
efli
es to m
ov
e
towa
rd the
othe
r
fi
ref
ly
.
3.
THE
D
IF
FER
ENTIAL E
V
OLUTIO
N (D
E) A
L
GO
RIT
HM
DE
al
gorithm
is
one
of
the
powe
rful
e
vo
l
ut
ion
a
ry
al
gorith
m
s
to
so
lve
re
al
par
am
et
er
optim
iz
at
ion
pro
blem
s.
Seve
ral
sta
te
-
of
-
t
he
-
art
e
voluti
on
aryal
gorithm
s
us
e
DE
as
e
voluti
onary
m
echan
ism
[15
]
-
[
19]
.
The
DE
al
gor
it
h
m
is
a
popula
ti
on
base
d
al
gorithm
li
ke
ge
netic
al
gor
it
h
m
s
us
in
g
th
e
si
m
il
ar
op
e
r
at
or
s;
cro
ss
over
,
m
utati
on
an
d
sel
ect
ion
.
T
he
m
ajor
c
on
tr
a
st
in
buil
ding
be
tt
er
arr
a
ng
em
ents
is
that
ge
netic
est
i
m
at
es
dep
end
on
c
om
bin
at
ion
w
hile
DE
dep
e
nds
on
co
nv
e
rsion
ope
ra
ti
on
.
T
his
m
aj
or
op
e
rati
on
de
pend
s
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Linear P
hase
FIR Low
Pass
Fil
te
r D
esi
gn
Base
d on
..
.
(
M
oa
t
h Sab
abha
)
4359
on
the
disti
nct
ion
s
of
ar
bitra
rily
te
ste
d
set
s
of
m
et
ho
ds
i
n
the
popula
ce.
As
a
r
ule,
DE
al
gorithm
has
three
sta
ges
[15]:
a.
Muta
ti
on
:
At generati
on
k for
each
par
e
nt
ve
ct
or
i
k
u
is ge
ner
at
e
d
acc
ordin
g
t
o fo
ll
owin
g
e
qua
ti
on
)
(
3
2
1
r
k
r
k
r
k
i
k
x
x
F
x
u
(8)
Wh
e
re
i
r
r
r
3
2
1
an
d
F
is r
efer
red as the
s
cal
ing
facto
r.
b.
Cros
s
over:
T
hi
s
operat
or
is
e
ssentia
l
due
to
it
s
aff
ect
abili
ty
to
co
nfrontdi
sti
nct
and
no
n
-
disti
nguish
a
bl
e
pro
blem
s.
A
tria
l
child
vecto
rs
]
,..
..,
,
[
2
1
i
nk
i
k
i
k
i
k
y
y
y
y
are
de
velo
pe
d
by
the
hy
br
i
d
of
it
s
par
e
nt
vec
tor
i
k
x
and it
s m
utant v
ect
or
i
k
u
as f
ollo
ws:
,
l
j
and
C
U
if
x
l
j
or
C
U
if
u
y
r
j
i
jk
r
j
i
k
i
jk
(9)
In
t
he
ab
ove
m
od
el
,
l
is
anar
bitra
rily
pick
ed
inte
ger
fro
m
{
1
;
2
;…
;
d},U
1
;U
2
;…
;Ud
are
a
r
bitrary
ind
e
pende
nt
va
riables
un
i
for
m
ly
disse
m
ina
te
d
in
[0
;
1),
a
nd
C
r
€
[0
;
1]
is
an
i
nput
pa
ram
et
er
aff
ect
ing
th
e
nu
m
ber
of com
pone
nts to
b
e
e
xch
a
nged
b
y t
he
cr
os
s
ov
e
r.
c.
Sele
ct
ion
(
Determ
inati
on
):
A
gr
ee
dy
Sele
ct
ion
em
plo
ye
d
wh
e
re
a
child
su
bs
ti
tutes
it
s
par
e
nt
in
ne
xt
gen
e
rati
on if
it
h
as a
supe
rior
or eq
uiv
al
e
nt f
i
tness
functi
o
n value.
In
[
19]
,
they
pro
po
se
d
the
hybri
d
DE
al
gorithm
with
local
searc
h
te
chn
i
qu
e
s
an
d
an
ada
ptive
par
am
et
er v
al
ue
f
or
C
r.
This a
dap
ti
ve
m
echan
ism
co
m
bin
es the b
i
nar
y c
rosso
ver
a
nd li
ne
r
ecom
bin
at
io
n. A
nd,
refreshm
ent m
echan
ism
is u
s
ed
to
avoid
sta
gn
at
io
n.
The
m
ajo
r
step
s of the
DE are
[19
]
:
1.
S1
:
In
it
ia
li
ze P(0) wit
h N
p
in
di
vid
ual
ra
ndoml
y sel
ect
ed
fro
m
the searc
hing s
pace.
2.
S2
: E
valuate i
ni
ti
al
p
opulati
on P(0).
3.
rep
eat
4.
S3
: F
or
k
=
1 :
Ma
xG
e
n or Co
nv
e
r
gen
ce
Crit
erio
n
reac
he
d
a.
Sele
ct
r
an
dom
l
y a su
bpop
ulati
on of
N
s
(
k) in
di
vid
uals
w
it
h p
rop
os
ed
so
l
utio
ns
on P
(k).
b.
Apply t
he f
undam
ental
D
E
operator “ M
utati
on and C
ross
over”
to get t
he
offsprin
g O
(k)
c.
Fo
r
e
ver
y c
hild
, p
e
rfo
rm
a local search
.
d.
Evaluate
offs
pri
ng
O
(k): if c
hi
ld
par
e
nt,
the
parent is
substi
tuted
by it
s c
hild
e.
If
i
nter
qu
a
rtil
e range
(IQR)
<
^
V
;
upd
at
e t
he p
opulati
on
.
5.
un
ti
l st
op
ping
conditi
on
;
6.
Algorithm
En
d
4.
THE
PARTI
CLE SW
A
R
M
O
PTIMIZ
ATIO
N
Partic
le
Sw
ar
m
Op
tim
iz
ation
(P
S
O
)
is
an
ev
olu
ti
onary
al
go
rithm
presented
a
nd
de
sign
e
d
by
Kenne
dy
an
d
Eberha
rt
in
19
95
[
20
]
.
S
om
e
effor
ts
ha
ve
be
en
prese
nted
towa
rd
t
he
ad
va
ncem
ent
of
th
e
FIR
Fil
te
r
base
d
on
the
PS
O
al
gor
it
h
m
.
The
PSO
easi
ly
app
li
cable.
Its
m
erg
in
g
m
igh
t
be
overseen
as
us
i
ng
few
var
ia
bles.
P
S
O
is
a
ver
sat
il
e,
incre
d
ible
popula
ce
based
stochastic
pursu
it
or
e
nhan
ce
m
ent
strat
eg
y
with
com
pr
ehe
nd
e
d
par
al
le
li
sm
,
wh
ic
h
ca
n,
with
no
i
ncon
ve
nience,
handle
non
-
dif
fer
e
ntial
reason
capaci
ti
e
s.
PS
O
is
le
ss
vulne
r
able
to
get
t
r
app
e
d
on
rest
rict
ed
op
ti
m
a
dissim
il
ar
Gen
et
ic
Al
gorit
hm
(G
A
),
re
pl
ic
at
ed
Anneali
ng,
et
c
.
in
[
22
-
23
]
,
t
he
y
exp
a
nd
e
d
a
si
m
il
ar
PSO
idea
to
pr
ese
nt
a
swar
m
of
bir
ds
.
P
SO
is
e
xp
and
e
d
thr
oughout
re
pl
ic
at
ion
of
bir
d
floc
king
i
n
m
ulti
di
m
ension
a
l
sp
ace.
Acc
or
ding
to
[
24]
,
fl
ock
i
ng
bi
rd
s
c
an
be
us
e
d
to
op
ti
m
i
ze
a
cer
ta
in
obj
ect
ive
f
un
ct
io
n.
I
n
PS
O,
eve
ry
po
ssi
ble
so
l
ution
is
de
note
d
as
a
par
ti
cl
e.
Every
par
ti
cl
e
is
ass
oc
ia
te
d
with
a
posit
ion
x
an
d
a
velocit
y
v.
on
e
can
ex
pr
e
ss
t
he
posit
ion
a
nd
velocit
y
of
t
he
it
h
par
ti
cl
e as
fo
ll
ow
i
ng
:
,
1
,
2
,
(
,
,
.
.
.
,
)
i
i
i
i
N
x
x
x
x
(10)
,
1
,
2
,
(
,
,
.
.
.
,
)
i
i
i
i
N
v
v
v
v
(11)
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
87
08
In
t J
Elec
&
C
om
p
En
g,
V
ol.
8
, N
o.
6
,
Dece
m
ber
2
01
8
:
4356
-
4365
4360
The
le
ngth
of
ever
y
vecto
r
N
de
note
s
th
e
dim
ension
of
t
he
pro
ble
m
or
num
ber
of
unkn
own
var
ia
bles.
In
e
ach
it
erati
on,
t
he
c
os
t
f
unct
ion
is
pr
ocesse
d
f
or
e
ver
y
one
of
pa
rtic
le
s
in
the
swa
rm
.
This
capaci
ty
ou
ght
to
be
pr
eci
se
ly
design
ed
to
giv
e
a
ref
le
ct
ion
of
the
ex
pe
ct
ed
ou
tc
om
e
.
The
posit
ion
and
velocit
y
of
eac
h
pa
rtic
le
are
updated
acc
ordin
g
to
in
div
id
ual
best,
xbest
i,
and
global
be
st,
xb
est
,
fitn
ess
as
sh
ow
n belo
w:
),
(
)
(
2
2
1
1
1
i
b
e
s
t
n
i
b
e
s
t
i
n
n
i
n
n
i
x
x
c
x
x
c
v
v
1
1
n
i
i
n
i
tv
x
X
(12)
The
s
upersc
rip
ts
n+1
an
d
n
i
ndic
at
e
the
ti
m
e
recor
d
of
t
he
current
a
nd
the
past
cy
cl
es
a
nd
β
1
a
nd
β
2
are
arb
it
ra
ry
num
ber
s
that
are
co
ns
ist
ently
ci
rcu
la
te
d
in
the
interim
(
0,
1).
T
hese
r
andom
nu
m
be
rs
are
refreshe
d
eac
h
tim
e
they
happ
en.
Th
e
com
pa
rati
ve
wei
gh
ts
of
the
pe
rsonal
best
posit
ion
c
orres
pondin
g
to
the
global be
st p
osi
ti
on
are dete
r
m
ined
by t
he p
aram
et
ers
c1
a
nd c2, re
sp
ect
i
vely
. Bo
t
h
c1
a
nd c2 are ty
pic
al
ly
se
t
to
a
value
of
2.
0.
T
he
pa
ram
eter
γ
n
is
t
he
“i
ner
ti
a
w
ei
ght”
in
the
nth
it
era
ti
on
.
It
is
a
num
ber
in
the
ra
ng
e
(
0,
1)
that
sel
ect
s
the
weigh
t
by
wh
ic
h
the
pa
rtic
le
’s
curre
nt
velocit
y
reli
es
on
it
s
pr
e
vi
ou
s
velocit
y,
and
th
e
distance
betwe
en
the
p
a
rtic
le
’
s posit
ion an
d
i
ts perso
nal
bes
t and gl
ob
al
b
e
st p
os
it
ion
s
.
The
po
pu
la
ti
on
of
par
ti
cl
es
is
then
m
ov
ed
accor
ding
to
e
qu
at
io
n
(
12)
a
nd
te
nds
to
cl
ust
er
tog
et
he
r
from
diff
eren
t
directi
ons.
Sin
ce,
a
m
axi
m
u
m
velocit
y,
Vma
x,
sh
ould not
increase
wit
h
any
par
ti
cl
e
to
sta
y
the
search
withi
n
a
desire
d
so
l
ution
sp
ace
.
Asse
ssm
ent
of
the
c
os
t
functi
on
is
done
util
iz
ing
the
pa
rtic
le
’s
ne
w
po
sit
io
n.
T
he
al
gorithm
go
es
throu
gh
the
se
procedu
res
it
erati
vely
un
ti
l
the
po
int
that
a
par
ti
cul
ar
en
d
par
a
dig
m
is m
et
.
5.
THE
THE FI
REFLY
ALG
ORI
TH
M
In
[
26
]
,
there
i
s
a
cl
arifica
ti
on
of
ho
w
the
F
irefly
al
go
rith
m
that
fo
ll
ows
the
fire
fly
be
ha
vior.
Firefly
is
an
insect
t
hat
f
or
t
he
m
os
t
pa
rt
deliv
ers
s
hort
an
d
caden
ce
d
flas
hes
that
c
reated
by
a
proce
dure
of
bio
lum
inescen
ce.
The
capaci
t
y
of
the
glim
merin
g
li
gh
t
is
to
pull
in
acco
m
pl
ic
es
(co
r
re
sp
on
de
nce)
or
dr
a
w
in
po
te
ntial
pr
ey
and
as
a
de
fensi
ve
cauti
onin
g
toward
the
pre
dato
r.
Al
ong
these
li
nes,
this
power
of
li
gh
t
is
the
factor o
f othe
r firefli
es in
a
dvancin
g
to
wa
rd
the o
t
her fire
fl
y.
The
li
gh
t
pow
er
is
chan
ge
d
at
the
separ
at
ion
f
r
om
the
eyes
of
the
onlo
ok
e
r
.
It
is
protect
ed
to
sta
te
that
the
li
ght
powe
r
is
dim
inishe
d
as
the
dis
ta
nce
incr
em
e
nt.
T
he
li
ght
powe
r
li
ke
wise
the
im
pact
of
the
ai
r
retai
ns
by
the
env
i
ronm
ent,
in
this
way
t
he
f
or
ce
t
urns
ou
t
to
be
le
ss
e
ngagi
ng
as
t
he
separ
at
io
n
incr
e
m
ent.
Firefly
al
g
or
it
hm
or
iginall
yp
resen
te
d
base
d
on
th
ree
idea
li
ze
ru
le
s,
1)
Fireflie
s
are
at
tract
ed
to
ward
each
oth
e
r’
s
re
gardl
ess
of
ge
nd
e
r.
2)
Th
e
e
ng
a
gi
ng
qual
it
y
of
t
he
firef
li
es
is
correla
ti
ve
with
th
e
s
plen
dor
of
the
firef
li
es.
C
on
se
qu
e
ntly
,
the
le
ss
app
eal
i
ng
fir
e
fly
will
pu
s
h
ahead
t
o
the
m
or
e
al
lu
rin
g
fir
efly
.
3)
T
he
s
hi
ne
of
firef
li
es is
rely
ingu
pon
t
he
c
ost
f
unct
io
n
[
27
-
30
]
.
5.1.
St
ru
cture
of
Fir
efly
A
l
gori
th
m
In
firef
ly
al
go
rithm
,
there
ar
e
two
esse
ntial
factor
s,
w
hich
is
the
li
gh
t
i
ntensity
force
and
a
ppeal
.
Firefly
is
pu
ll
ed
in
to
ward
the
oth
e
r
fire
f
ly
that
has
br
igh
te
r
blaze
th
an
it
sel
f.
The
eng
a
ging
qua
li
ty
is
dep
e
nded
with
the
li
gh
t
pow
er.
T
he
li
ght
intensit
y
acco
r
dingly
at
tract
ivene
ss
is
in
versel
y
relat
ive
w
it
h
the
par
ti
cula
r
distance
from
the
l
igh
t
sour
ce.
In
t
his
m
ann
er
,
th
e
li
gh
t
a
nd
e
ng
agin
g
qual
it
yi
s
dim
inishing
a
s
the
distance i
ncr
e
m
ent. One ca
n express
it
as fo
ll
ow
s:
(
)
=
0
−
2
(13)
wh
e
re,
I
=
li
gh
t i
nte
ns
it
y,
0
=
li
gh
t i
nte
ns
it
y at
init
ia
l or
or
i
gin
al
li
ght
intensit
y,
=
the li
gh
t
abs
or
ption coe
ff
ic
ie
nt
r
=
distance
betwe
en firefly
i
a
nd
j
Attract
ivene
ss
is
propo
rtiona
ll
y
to
the
li
ght
inte
ns
it
y
se
en
by
a
no
t
her
fire
flie
s,
t
hu
s
at
tract
ivenes
s
(
)
is
giv
e
n by
=
0
−
2
(14)
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Linear P
hase
FIR Low
Pass
Fil
te
r D
esi
gn
Base
d on
..
.
(
M
oa
t
h Sab
abha
)
4361
wh
e
re
0
re
pr
ese
nts the
att
racti
ven
e
ss at ze
ro
distance
(
=
0
)
. T
he dist
ance
betwe
en
tw
o fire
flie
s can be
def
i
ned b
ase
d on the Ca
rtesi
an dist
ance as
foll
ows:
=
|
−
|
=
√
∑
(
,
−
,
)
2
=
1
(15)
Firefly
i
is at
tr
act
ed
to
ward t
he
m
or
e att
ract
ive f
ire
fly
j
, the
upd
at
e
d
m
ove
m
ent is ex
pr
e
ssed
as
foll
ows
:
∆
=
0
−
2
(
−
)
+
,
+
1
+
∆
(16)
wh
e
re
0
is
f
or
a
tt
racti
on
,
is
th
e
lim
it
ation
w
hen
the
valuet
ends
t
o
ze
r
o
o
r
to
o
la
rg
e
.
If
a
ppr
oach
i
ng
ze
r
o
(
→
0
)
,
the
at
tract
ive
ness
a
nd
br
ig
ht
ness
bec
om
e
c
on
sta
nt,
=
0
.
I
n
a
nothe
r
w
ord
,
a
f
irefly
can
be
s
een
in
any
po
sit
io
n,
easy
to
c
om
plete
global
se
arch.
If
the
is
nea
rin
g
i
nfi
nity
or
to
o
l
arg
e
(
→
∞
)
,
t
he
at
tract
iveness
and
bri
ghtness
beco
m
e
decr
e
ase.
The
fire
fly
m
ov
em
ents
beco
m
e
random
.
The
i
m
ple
m
entat
io
n
of
firef
ly
al
go
rithm
can
be
done
in
t
hese
two
asy
m
pto
ti
c
beh
a
viors.
Wh
il
e
the
sec
ond
the
te
rm
is
f
or
rand
om
iz
a
ti
on
,
as
is
the
rando
m
iz
epar
am
e
te
r.
T
he
can
be
re
placed
by
ran
-
1/2
w
hich
is
ra
n
is
ra
nd
om
nu
m
ber
ge
ner
a
te
d
f
ro
m
0
to
1. Please
ref
e
r
t
o
Algorithm
1
.
Al
go
rit
hm
1:
Fi
refly algorit
h
m
Inp
ut
:
Co
st F
unct
ion
(
)
, Initi
al p
op
ula
ti
on
of
Fire
fl
ie
s
0
,
Defi
ne
the
li
gh
t absor
ptio
n
c
oe
ff
ic
ie
nt
,
Max
num
ber of
Iteratio
n
, in
it
iali
ze the
Lig
ht
In
te
ns
it
y
0
at
0
by
(
)
Whi
le
Lo
op
:
till
≤
For lo
op:
for e
ach
=
1
…
all
fi
ref
lies
Inner l
oop:
for
ea
c
h
=
1
…
all
fi
ref
l
ie
s
If
(
<
)
, move fire
fl
y
tow
ar
ds
;
en
d
if
.
Varying
a
tt
r
ac
ti
veness
wi
th di
stan
ce
due t
o
−
Ev
alua
te
new
so
lut
ions
an
d upd
ate the li
ght i
ntensity
End
F
or
End
F
or
Rank the
firefl
ie
s
and f
ind t
he
cu
rre
nt
globa
l
b
est
v
alu
es
End
Whi
le
Ou
tp
ut:
t
he be
st f
ire
fl
ie
s so
lu
ti
on
s
6.
SIMULATI
O
N RESULTS
In
t
his
sect
io
n,
we
s
how
t
he
util
iz
at
ion
of
t
he
fi
ref
ly
al
go
rithm
to
desig
n
op
ti
m
u
m
FI
R
filt
ers
f
or
diff
e
re
nt
cases.
In
al
l
cases, f
or
the
sa
ke
of
sim
pl
ic
it
y,
the
filt
er
to
be
de
sig
ned
is
ass
um
ed
to
be
a
li
near
p
ha
s
e
LPF
with
eve
n
le
ng
th
to
co
nst
ru
ct
just
50
%
of
the
coe
ff
ic
i
ents
rathe
r
than
al
l
coef
fici
en
ts.
Ph
ase
li
neari
ty
of
the
filt
er
is
guaran
te
e
d
by
as
su
m
ing
sy
m
m
et
ry
of
the
a
pproxim
at
e
filt
er
wh
ic
h
re
du
ce
s
the
dim
ension
of
th
e
op
ti
m
iz
ation
pro
blem
to
M/
2.
In
t
he
fir
st
case,
the
firef
ly
al
gorithm
is
us
ed
to
desig
n
optim
u
m
FI
R
filt
ers
in
wh
ic
h
M
,
w
p,
w
s,
and
the
rati
o
δ
p
/δ
s
are
fixed.
For
th
is
case,
the
fitness
functi
on
to
be
m
ini
m
i
zed
is
def
i
ned as:
c
o
s
m
a
x
(
|
(
)
|)
,
ps
t
w
F
F
F
E
w
(17)
Wh
e
re
F
p
an
d
F
s
are
the
cl
ose
d
subsets
0
≤
w
≤
w
p
and
w
s
≤
w
≤
0
.
5,
res
pe
ct
ively
.
Figu
r
e
1
pr
ese
nt
s
the
fr
e
qu
e
ncy
respo
ns
e
of
an
app
r
oxim
at
e
filt
er
with
M
=
30,
w
p
=
0
.
25,
w
s
=
0
.
3,
δ
p
/δ
s
=
1,
based
on
the
Firefly
al
gorithm
,
DE,
PSO
a
nd
PM
al
gorithm
s,
on
the
sam
e
gr
a
ph.
T
he
fi
gure
shows
t
hat
th
e
Firefly
res
po
ns
e
is
pr
eci
sel
y
giv
i
ng
the
sta
ndar
d
FI
R
optim
al
desig
n.
T
his
sign
i
fies
that
the
firef
ly
al
gorithm
is
per
for
m
s
well
al
ong
the
PSO
and
DE
al
gori
thm
in
te
r
m
s
of
accuracy.
I
n
this
case,
the
firef
ly
,
DE
an
d
PSO
we
re
ra
ndom
ly
init
ia
li
zed an
d
conve
rg
e
d wit
hin
a
r
eas
onabl
e tim
e.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
87
08
In
t J
Elec
&
C
om
p
En
g,
V
ol.
8
, N
o.
6
,
Dece
m
ber
2
01
8
:
4356
-
4365
4362
Fig
ure
1. Fr
e
quency
res
pons
e
of linear
phase
FI
R
filt
er for
M
=
30,
wp
=
0.2
, w
s
=
0.2
5,
δ
p/δs
=
1
Fo
r
set
tl
ed
pa
r
a
m
et
ers
case,
the
fire
fly
al
go
rithm
is
utilized
to
desig
n
a
l
inear
ph
a
se
FIR
filt
ers
in
wh
ic
h
M
,
w
p,
w
s,
δ
p,
an
d
δ
s
are
ch
os
e
n
by
the
desig
ner.
In
Fi
gure
2,
t
he
de
sig
n
proc
ess
be
gin
s
with
the
desire
d
filt
er
pa
ram
et
ers
(
M
,
w
p,
w
s
)
,
and
t
he
PM
al
go
rit
hm
is
e
m
plo
yed
to
ob
ta
i
n
the
filt
er
c
oeffici
ents.
The
n,
the
desi
gn
e
r
sel
ect
s
a
f
easi
ble
value
f
or
δ
p
a
nd
δ
s
ba
sed
on
the
m
axim
u
m
ripp
le
siz
e
(
δ
PM
)
obta
ined
by
the
PM.To
ha
ve
the
capaci
ty
to
con
tr
ol
the
r
ipp
le
s
in
the
two
bands
in
de
pende
ntly
,
on
e
can
re
-
w
rite
the
cost
functi
on in
(17) as
f
ollow
s:
c
os
m
a
x
(
|
(
)
|
)
m
a
x
(
|
(
)
|
)
ps
t
p
s
w
F
w
F
F
E
w
E
w
(18)
Figure
2
sho
ws
the
fr
e
quency
respo
ns
e
of
a
li
near
phase
FI
R
desi
gn
e
d
base
d
on
the
fire
fly
al
gorithm
in
wh
ic
h
the
f
il
te
r
par
am
et
er
s
are
set
to
M
=
30,
w
p
=
0
.
25,
w
s
=
0
.
3,
δ
p
=
0
.
1,
δs
=0
.
01.
Th
e
PM
resp
ons
e
for
M
=
30,
w
p
=
0
.
25,
w
s
=
0
.
3,
a
nd
δ
p
/δ
s
=
10
0
is
al
so
s
how
n
in
th
e
sa
m
e
figu
re
. O
bvio
us
ly
,
F
ig
ur
e 2
sho
ws
th
at
the
PM
al
gorithm
do
e
s
no
t
al
lo
w
to
c
ontrol
the
pass
ba
nd
a
nd
stopba
nd
rip
pl
es.
Howe
ver
,
t
he
Fi
ref
ly
al
go
rith
m
can
co
ntr
ol
th
ese
qu
al
it
ie
s
of
rip
ple
in
the
pass
band
an
d
stopban
d
sim
ultaneo
us
ly
.
T
he
fire
fly
al
go
rithm
achieves
to
t
he
d
esi
re
d rip
ples
in bo
t
h bands
.
Fig
ure
2. Fr
e
quency
res
pons
e
of linear
phase
FI
R
filt
er, for
M
=
30
,
wp
=
0.
25, ws
=
0.3, δ
p
=
0.1, δs=
0.01
0
0
.
0
5
0
.
1
0
.
1
5
0
.
2
0
.
2
5
0
.
3
0
.
3
5
0
.
4
0
.
4
5
0
.
5
-
8
0
-
7
0
-
6
0
-
5
0
-
4
0
-
3
0
-
2
0
-
1
0
0
10
N
o
r
m
a
l
i
z
e
d
f
r
e
q
u
e
n
c
y
M
a
g
n
i
t
u
d
e
i
n
d
B
M
a
g
n
i
t
u
d
e
r
e
s
p
o
n
s
e
o
f
t
h
e
e
q
u
i
r
i
p
p
l
e
,
l
i
n
e
a
r
p
h
a
s
e
FI
R
f
i
l
t
e
r
f
i
r
Fi
r
e
f
l
y
D
e
s
i
g
n
f
i
r
P
S
O
D
e
s
i
g
n
f
i
r
D
E
D
e
s
i
g
n
f
i
r
p
m
D
e
s
i
g
n
0
0
.
0
5
0
.
1
0
.
1
5
0
.
2
0
.
2
5
0
.
3
0
.
3
5
0
.
4
0
.
4
5
0
.
5
-
9
0
-
8
0
-
7
0
-
6
0
-
5
0
-
4
0
-
3
0
-
2
0
-
1
0
0
10
N
o
r
m
a
l
i
z
e
d
f
r
e
q
u
e
n
c
y
M
a
g
n
i
t
u
d
e
i
n
d
B
M
a
g
n
i
t
u
d
e
r
e
s
p
o
n
s
e
o
f
t
h
e
e
q
u
i
r
i
p
p
l
e
,
l
i
n
e
a
r
p
h
a
s
e
FI
R
f
i
l
t
e
r
f
i
r
Fi
r
e
f
l
y
D
e
s
i
g
n
f
i
r
p
m
D
e
s
i
g
n
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Linear P
hase
FIR Low
Pass
Fil
te
r D
esi
gn
Base
d on
..
.
(
M
oa
t
h Sab
abha
)
4363
It
is
un
reali
sti
c
to
acknow
le
dge
FI
R
cha
nne
ls
witho
ut
c
oeffici
ents
qu
a
nti
zat
ion
util
iz
ing
finite
wo
r
d
le
ng
th
.
This
outl
ine
require
m
ent
has
no
t
been
c
on
si
der
e
d
in
the
pr
i
or
work
or
e
ven
in
the
PM
al
go
rithm
.
Con
si
der
i
ng
th
e
lim
it
ation
s
in
filt
er
i
m
plem
entat
ion
,
the
su
it
able
word
le
ng
th
re
duce
al
go
rithm
com
plexit
y
and
eq
uip
m
ent
pr
e
requisi
te
s
can
prom
pts
a
dram
atic
change
i
n
the
fr
e
quency
re
sp
onse
an
d
de
sign
sp
eci
ficat
io
n
of
the
desire
d
F
IR.
H
ow
e
ver,
the
capaci
ty
of
firef
ly
al
gorith
m
to
deal
with
this
so
rt
of
iss
ues
is
exh
i
bited
in
th
e
accom
pan
yi
ng
cases.
First,
t
he
i
m
pact
of
the
quantiz
in
g
process
f
or
the
filt
er
coef
fici
e
nts
on
the
fr
e
qu
e
ncy
respo
ns
e
of
filt
ers
desig
ned
base
d
on
the
PM
is
de
m
on
s
trat
ed
in
Fig
ur
e
3.
In
this
case,
the
coeffic
ie
nts
ac
qu
i
red
base
d
on
the
PM
m
et
h
od
a
re
rou
nd
e
d
by
ei
ght
bits
word
le
ngth.
A
s
show
n
in
F
ig
ur
e
3,
the
re
spo
ns
e
of
the
PM
filt
er
with
8
-
bits
quantiz
ed
c
oeffici
ents
cha
nges
an
d
it
s’
pe
r
form
ance
dim
ini
sh
es
com
par
ing
to
t
he
fl
oatin
g
-
po
i
nt
PM
case.
I
n
F
i
gure
4,
w
e
pr
e
sent
t
he
r
esult
of
the
pr
esented
al
gorithm
to
const
ru
ct
t
he
optim
u
m
FI
R
fi
lt
er
for
w
hich
their
c
oeffici
ents
are
r
ounde
d
us
in
g
8
bits
word
le
ngth
f
or
th
e
sam
e fil
te
r
in
previ
ou
s
case.
Fig
ure
3.
Ef
fec
t of r
oundin
g o
n
F
IR f
il
te
rs
d
e
sign
e
d by PM
Figure
4.
FI
R
fi
lt
er d
esi
gne
d usin
g DE
with
coeffic
ie
nts
quantiz
ed usin
g 8
b
it
s
word le
ng
th for
M
=
30,
wp
=
0.2
5,
ws
=
0.3, δ
p/δs
=
1
Cl
early
,
the
Fi
ref
ly
perform
a
nce
is
bette
r
th
an
t
he
on
e
ac
quire
d
from
qu
a
ntizi
ng
of
t
he
coeffic
ie
nts
of
t
he
PM
m
eth
od.
It
is
wort
h
to
stu
dy
the
i
m
pact
of
the
order
of
the
FIR
filt
er
on
the
pr
ese
nted
m
eth
od.
I
n
ano
t
her
e
xam
ple,
the
Firefl
y
al
go
rithm
is
e
m
plo
ye
d
to
desig
n
a
n
F
I
R
filt
er
with
sp
eci
ficat
io
ns
M
=
15
,
wp
=
0.2
5,
ws
=
0.3,
δ
p/δs
=
1
i
n
wh
ic
h
the
c
oe
ff
ic
ie
nts
ar
e
r
ounde
d
usi
ng
8
bits
w
ord
le
ngth.
Fi
gure
5
pre
sents
the
pe
rfo
rm
ance
of
usi
ng
the
pr
ese
nted
m
et
h
od
to
c
onstr
uct
the
op
ti
m
u
m
filt
er
for
w
hic
h
the
coe
ff
ic
ie
nts
are
qu
a
ntize
d
usi
ng
ei
gh
t
bits
w
ord
le
ngth
.
Obv
iou
sly
,
the
Fire
fly
m
et
ho
d
pr
e
form
s
well
and
bette
r
than
the
on
e
acqu
i
red f
ro
m
r
ou
nd
i
ng of th
e coe
ff
ic
ie
nts
of the
PM m
et
ho
d wit
h sm
al
ler
order fil
te
r.
0
0
.
0
5
0
.
1
0
.
1
5
0
.
2
0
.
2
5
0
.
3
0
.
3
5
0
.
4
0
.
4
5
0
.
5
-
9
0
-
8
0
-
7
0
-
6
0
-
5
0
-
4
0
-
3
0
-
2
0
-
1
0
0
10
M
a
g
n
i
t
u
d
e
r
e
s
p
o
n
s
e
o
f
t
h
e
e
q
u
i
r
i
p
p
l
e
,
l
i
n
e
a
r
p
h
a
s
e
FI
R
f
i
l
t
e
r
M
a
g
n
i
t
u
d
e
i
n
d
B
N
o
r
m
a
l
i
z
e
d
f
r
e
q
u
e
n
c
y
f
i
r
p
m
D
e
s
i
g
n
r
o
u
n
d
i
n
g
0
0
.
0
5
0
.
1
0
.
1
5
0
.
2
0
.
2
5
0
.
3
0
.
3
5
0
.
4
0
.
4
5
0
.
5
-
1
0
0
-
8
0
-
6
0
-
4
0
-
2
0
0
20
M
a
g
n
i
t
u
d
e
r
e
s
p
o
n
s
e
o
f
t
h
e
e
q
u
i
r
i
p
p
l
e
,
l
i
n
e
a
r
p
h
a
s
e
FI
R
f
i
l
t
e
r
M
a
g
n
i
t
u
d
e
i
n
d
B
N
o
r
m
a
l
i
z
e
d
f
r
e
q
u
e
n
c
y
f
i
r
Fi
r
e
f
l
y
D
e
s
i
g
n
r
o
u
n
d
i
n
g
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
87
08
In
t J
Elec
&
C
om
p
En
g,
V
ol.
8
, N
o.
6
,
Dece
m
ber
2
01
8
:
4356
-
4365
4364
In
a
no
t
her
e
xa
m
ple,
we
incr
e
ase
the
qua
ntiza
ti
on
bits
to
equ
al
10
bits
.
The
firef
ly
a
lgorit
hm
i
s
e
m
plo
ye
d
t
o
desig
n
a
n
FIR
filt
er
wit
h
sp
eci
ficat
io
ns
M
=
30
,
wp
=
0.25,
ws
=
0.3,
δ
p/δs
=
1
in
wh
i
ch
t
he
coeffic
ie
nts ar
e
r
ou
nd
e
d by 10
b
it
s word l
en
gt
h.
T
he
res
ults
dep
ic
te
d i
n
F
i
gure
6
s
how
tha
t t
he
Firefly
m
et
hod
get an en
ha
nce
m
ent o
n t
he
p
e
rfor
m
ance and
sti
ll
o
utp
er
f
orm
s the r
oundin
g respo
ns
e
of P
M m
e
tho
d.
Fig
ure
5.
FI
R
fi
lt
er d
esi
gne
d usin
g DE
with
coeffic
ie
nts
quantiz
ed usin
g 8
b
it
s
word le
ng
th for
M
=
15,
wp
=
0.2
5,
ws
=
0.3, δ
p/δs
=
1.
Fig
ure
6.
FI
R
fi
lt
er d
esi
gne
d usin
g DE
with
coeffic
ie
nts
quantiz
ed usin
g 1
0 bit
s word le
ngth
f
or
M
=
30,
wp
=
0.2
5,
ws
=
0.3, δ
p/δs
=
1
7.
CONCL
US
I
O
N
Desig
ning
of
l
inear
ph
ase
FIR
filt
er
has
be
en
car
ried
ou
t
thr
ough
ap
plica
ti
on
of
fir
efly
al
go
rit
hm
.
The
F
IR
was
c
om
par
ed
with
the
PS
O
a
nd
DE
al
go
rithm
s
.
The
fire
fly
al
gorithm
sh
ows
the
a
bili
ty
to
desig
n
su
c
h
filt
ers
bot
h
in
fi
nite
and
infin
it
e
w
ord
l
eng
t
h
coe
ff
ic
ie
nts.
T
her
e
a
re
three
a
dv
a
ntag
es
ti
ed
to
the
f
irefly
al
gorithm
.
That
is
getti
ng
t
o
the
true
gl
obal
m
ini
m
u
m
reg
a
rd
le
ss
the
init
ia
l
par
am
et
er,
a
few
co
ntr
olled
par
am
et
ers
are
us
e
d
ye
t
t
her
e
is
fast
co
nver
gen
ce
.
T
he
res
ults
validat
e
th
e
fact
t
hat
it
i
s
possible
to
de
sig
n
op
ti
m
al
finite
word
le
ng
t
h
F
IR
filt
ers
by
use
of
fi
ref
ly
al
gorithm
.
Also
,
in
al
lc
ases,
the
fire
fly
al
gorith
m
perform
s
well
al
ong
with
t
he
DE
a
nd
PS
O
al
gorithm
s.
A
lt
ho
ug
h
s
oft
wa
re
im
ple
m
enta
ti
on
is
im
po
rtant
t
o
inv
est
igate
t
he
capa
bili
t
ie
s
of
Firefly
m
et
hod
a
nd
to
sim
ul
at
e
sign
i
ficant
aspects
of
F
IR
Fil
te
r
ap
plica
ti
on
s
,
hard
war
e
im
ple
m
entat
ion
pr
ov
i
des
r
eal
ti
m
e
so
luti
ons
a
nd
a
n
optim
al
par
al
le
li
sm
method
in
te
rm
s
of
fast
conve
rg
e
nce.
Hen
ce
,
ha
r
dw
a
re
im
ple
m
entation
s
ca
n
be
co
ns
ide
red
a
s
a
prom
isi
ng
ap
pro
ach
to
im
ple
m
ent
th
e
pr
ese
nted
m
et
ho
d w
hich
can
be exec
uted
b
y
I
nteg
rated Circ
uit (I
Cs
).
REFERE
NCE
S
[1]
L.
Li
twin
,
FIR
a
nd
IIR
dig
it
a
l
f
il
t
ers,
I
EE
E
Potential
s.
0278
-
6648,
2000,
28
–
31
.
[2]
Harpre
et
K
aur
and
Bal
wind
er
Dhali
wal
,
“
Design
of
low
pa
s
s
FIR
fil
te
r
using
art
ifici
al
n
eur
al
n
et
work
”
,
Inte
rna
ti
ona
l
Jou
rna
l
of
Inform
at
i
on
and El
ec
tron
i
cs
Engi
n
ee
rin
g,
Vol.
3
,
No.
2,
Marc
h
2013
.
[3]
Sonika
Gupta
,
Am
an
Panghal
,
“
Perform
anc
e
Anal
y
s
is
of
FIR
Filt
er
Design
b
y
Us
ing
Recta
ng
ula
r,
Hann
ing
a
nd
Ham
m
ing
W
indows
Methods
”
,
Inte
rna
ti
ona
l
Journal
of
Advanc
ed
Resea
r
ch
in
Com
pute
r
Scie
n
ce
and
Softwar
e
Engi
ne
eri
ng,
Vo
lume
2,
Iss
ue
6,
June
2012.
[4]
J.H.
McCle
llan,
T.
W
.
Parks,
L.
R
.
Rabi
ner
,
A
computer
progra
m
for
designi
ng
opti
m
um
FIR
li
nea
r
phase
digital
fil
ters
,
I
E
EE T
r
a
ns.
Audio
El
e
ct
r
oac
oust.
AU
-
21
(1973)
506
–
526.
[5]
L.
R.
Rabi
n
er,
A
pproximate
d
esign
rel
a
ti
onships
for
low
-
pass
FIR
digi
t
al
fi
lters,
IEE
E
Tra
ns.
Audio
Elec
tro
ac
ous
t.
AU
-
21
(1973)
4
56
–
460.
[6]
Z.
Albataine
h
,
F.
Sale
m
,
"N
ew
Bli
nd
Multi
use
r
Dete
ction
in
DS
-
CDM
A
U
sing
H
-
DE
and
I
CA
Algorit
hm
s"
,
Inte
rna
ti
ona
l
Co
nfe
ren
c
e
on
201
3
4th
In
te
l
li
g
ent
S
y
stems
Modell
i
ng
&
Sim
ulation
(ISM
S),
pp.
569
-
574,
Jan
.
2013
[7]
G.
Singh,
N.
Pra
kash,
“
FP
GA
Impl
ementation
of
Higher
Order
FI
R
Filt
er
”,
Inte
rn
at
ion
al
Journa
l
o
f
Elec
tr
ical
and
Com
pute
r
Engi
n
ee
ring
(IJECE)
,
Vol.
7
,
No.
4,
pp
.
1874~1881,
Au
gust
2017.
[8]
M.
Ali,
“
Casca
d
ed
Rippl
e
C
arr
y
Adder
bas
ed
S
RCS
A
for
Eff
ici
ent
FIR
Fil
te
r
”,
Indone
sian
Jour
nal
of
Elec
tri
c
al
Engi
ne
eri
ng
and
Com
pute
r
Sci
en
ce
,
Vol.
9,
No.
2
,
pp
.
253~256,
Februa
r
y
20
18
.
[9]
S.
Das
and
P.
Sugantha
n,
"P
roble
m
def
ini
t
ions
and
eva
lu
at
ion
cri
te
r
ia
for
c
e
c
2011
competition
on
te
stin
g
evol
uti
on
ar
y
a
lg
orit
hm
s
on
rea
l
world
opti
m
iz
ati
on
proble
m
s"
,
J
ada
vpur
univ
ersity
and
Nan
y
an
g
Te
chnol
og
ical
Univer
sit
y
,
T
ec
h
.
Rep
.
,
2011.
0
0
.
0
5
0
.
1
0
.
1
5
0
.
2
0
.
2
5
0
.
3
0
.
3
5
0
.
4
0
.
4
5
0
.
5
-
7
0
-
6
0
-
5
0
-
4
0
-
3
0
-
2
0
-
1
0
0
10
M
a
g
n
i
t
u
d
e
r
e
s
p
o
n
s
e
o
f
t
h
e
e
q
u
i
r
i
p
p
l
e
,
l
i
n
e
a
r
p
h
a
s
e
FI
R
f
i
l
t
e
r
M
a
g
n
i
t
u
d
e
i
n
d
B
N
o
r
m
a
l
i
z
e
d
f
r
e
q
u
e
n
c
y
f
i
r
Fi
r
e
f
l
y
D
e
s
i
g
n
r
o
u
n
d
i
n
g
0
0
.
0
5
0
.
1
0
.
1
5
0
.
2
0
.
2
5
0
.
3
0
.
3
5
0
.
4
0
.
4
5
0
.
5
-
8
0
-
7
0
-
6
0
-
5
0
-
4
0
-
3
0
-
2
0
-
1
0
0
10
M
a
g
n
i
t
u
d
e
r
e
s
p
o
n
s
e
o
f
t
h
e
e
q
u
i
r
i
p
p
l
e
,
l
i
n
e
a
r
p
h
a
s
e
FI
R
f
i
l
t
e
r
M
a
g
n
i
t
u
d
e
i
n
d
B
N
o
r
m
a
l
i
z
e
d
f
r
e
q
u
e
n
c
y
f
i
r
Fi
r
e
f
l
y
D
e
s
i
g
n
r
o
u
n
d
i
n
g
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Linear P
hase
FIR Low
Pass
Fil
te
r D
esi
gn
Base
d on
..
.
(
M
oa
t
h Sab
abha
)
4365
[10]
Gopinat
han
S.,
e
t
al.,
“
W
ave
let
a
nd
FF
T
Based
I
m
age
Denoising
Us
ing
Non
-
Li
n
ea
r
Fi
lt
ers,”
In
ternat
ion
al
Journa
l
of
Elec
tr
ical and
Com
pute
r
Eng
i
nee
ring
,
vo
l/
issue:
5(5)
,
pp
.
1018
-
1026,
2015
.
[11]
M.
Sharm
a,
Dr
S.K.
Singh,
“
New
Te
chni
qu
e
Based
Peasa
nt
Multi
plication
for
Eff
ci
en
t
Signal
Proce
ss
in
g
Applic
a
ti
ons”,
I
ndonesia
n
Journal
of
Elec
tr
ical
Engi
ne
eri
ng
and
Com
pute
r
Science
,
Vol
.
8,
No.
3,
pp.
726
~
729
Dec
ember
2017
.
[12]
T.
W
.
Parks,
J.H
.
McClella
n
,
Ch
eb
y
shev
appr
oxi
m
at
ion
for
non
-
rec
ursive
d
igi
t
al
fil
t
ers
with
li
n
ea
r
phase
,
I
EEE
Tra
ns.
C
irc
u
it
s T
heor
y
CT
-
19
(19
72)
189
–
194.
[13]
Z.
Alba
taineh,
F.
Sale
m
and
J.
I.
Ababne
h,
“
Li
n
ea
r
phase
FIR
L
ow
Pass
Filt
er
Design
Us
ing
Hy
brid
Diffe
r
ent
i
al
Evol
uti
on
”,
In
ternat
ion
al
Journal
of
Resea
rch
in
W
ire
le
ss
S
y
stems
(IJRWS),
Vol.
1,
Iss
ue
2,
pp.
43
-
49,
Dec
embe
r
2012.
[14]
Jeha
d
I.
Ababn
e
h,
Moham
m
ad
H.
Bat
a
ine
h
.
"Li
nea
r
phase
FIR
f
il
te
r
d
esign
usin
g
par
ticle
sw
arm
opti
m
iz
a
ti
on
an
d
gene
t
ic
al
gor
it
h
m
".
El
sevi
er,
Di
git
al Signa
l
Proc
essing,
2006
.
[15]
Q.
Zha
o,
G
.
Me
ng,
“
Design
of
Digit
al
FIR
Filt
e
rs
using
D
iffe
ren
ti
al
Evol
ut
ion
algorithm
base
d
on
Reser
ved
Gen
e”
,
Inte
rna
ti
ona
l
Co
nfe
ren
c
e
of
Infor
m
at
ion
Sci
ence and
Mana
g
emen
t
Eng
ine
er
ing
(I
SM
E),
pp.
177
-
1
80,
2010
.
[16]
Yang,
X.
and
K
ara
m
anogl
u,
M
.
,
Sw
arm
Inte
ll
igence
and
Bi
o
-
Inspire
d
Com
puta
tion.
Sw
arm
Inte
l
li
gen
ce
and
B
io
-
Inspire
d
Com
put
at
ion
,
pp
.
3
-
23,
2
013.
[17]
F.
Neri
and
V.
T
irrone
n,
"S
c
al
e
f
ac
tor
loc
a
l
sea
r
c
h
in
diffe
r
ent
i
al
evol
uti
on"
,
Me
m
et
ic
Com
puti
n
g,
vol.
1
,
pp.
15
3
-
171,
2009
,
10
.
10
07/s12293
-
009
-
0008
-
9.
[18]
S.
Das
and
P.
N.
Sugantha
n,
"D
iffe
ren
t
ia
l
evo
lut
i
on:
A
su
rve
y
of
the
stat
e
-
of
-
the
-
art
",
Evol
u
ti
ona
r
y
Com
puta
ti
on
,
IEE
E
Tr
ansa
ctio
ns on,
2010
.
[19]
Farook,
S.
and
Raj
u,
P.
(2013).
Evol
ut
iona
r
y
H
y
brid
Gene
t
ic
-
Fi
ref
l
y
Algori
thm
for
Global
Opt
i
m
iz
at
ion
.
IJCE
M
Inte
rna
ti
ona
l
Jou
rna
l
of
Com
putationa
l
Eng
ine
e
ri
ng
&
Man
age
m
e
n
t,
16(3)
,
pp
.
37
-
45.
[20]
J.
Kenne
d
y
,
R.
E
ber
har
t
,
’
"P
artic
l
e
Sw
ar
m
Optimiza
t
ion",
in
Proc
.
IEE
E
int.
Conf
.
On Neura
l
Netw
ork,
1995
[21]
Y.
Shi
and
R.
C.
Ebe
rha
r
t
(1998),
Para
m
et
er
Selec
ti
on
in
Particle
Sw
arm
Optimiz
at
ion
,
Evol
ut
ion
ar
y
Program
m
ing
VII,
Springer
,
L
ec
tu
r
e
Not
es
in
Com
pute
r
Scie
n
ce
1447
,
591
–
60
0.
[22]
S.
Das,
A.
Abraha
m
and
A.
Konar,
“
Parti
c
le
Sw
a
rm
Optimiza
ti
on
and
Diffe
ren
ti
a
l
Evol
uti
on
Algor
it
hm
s: T
ec
hni
cal
Anal
y
sis
,
Appli
ca
t
ions
and
H
y
bridi
z
at
ion
Pers
pec
t
ive
s”,
Studi
es
in
Com
puta
ti
onal
Intelli
g
ence
(SCI)
116,
1
-
38
(2008).
[23]
El
-
Saw
y
,
A.
,
Z
a
ki,
E
.
and
R
iz
k
-
Alla
h,
R
.
(2012)
.
A
Novel
H
y
bri
d
Ant
Colon
y
Optimiza
t
ion
and
Firefly
Algori
th
m
for
Solving
Con
strai
ned
Engi
n
eering
Design
Pro
ble
m
s.
Journal
o
f
Natur
a
l
Sci
ences
and
Ma
the
m
a
ti
cs,
6(1),
pp.
1
-
22.
[24]
D.
Manda
l,
S.
P
.
Ghos
hal
,
and
A
.
K.
Bhat
t
ac
h
arj
e
e,
“
A
Novel
Par
ti
cle
Sw
arm
Opt
imiza
ti
on
Based
Optimal
Design
of
Thre
e
-
Ring
Conce
ntr
i
c
Circular
Antenna
Ar
ra
y
,
”
IE
EE
Int
er
nat
ion
al
Confer
enc
e
on
Advanc
es
in
Com
puti
ng,
Control
,
and
Tel
ec
om
m
unic
at
ion
Technol
ogi
es,
(
ACT’09),
pp
.
38
5
-
389,
2009
.
[25]
Ab Wahab
MN
,
Neft
i
-
Mez
ia
ni
S,
At
y
abi A (
2015
)
A Com
pre
hens
ive
R
evi
ew
of
S
warm
Optimiza
tion Algori
thms
.
PLoS
ON
E
10(5):
e0122827
.
do
i
:10.
1371/journal
.
pone.
0122827
.
[26]
Yang,
X.
(2010). Nat
ure
-
Inspire
d
Meta
h
eur
isti
c
A
lgori
thms
.
2nd
e
d.
From
e:
Lu
niv
er
Press
.
[27]
Yu,
S., Yang, S. and
Su,
S.
(2013
).
Self
-
Ad
apt
iv
e S
te
p
Firef
l
y
Alg
orit
hm
.
Journal
of
Applie
d
Math
emati
cs,
pp.
1
-
8.
[28]
Ti
la
hun
,
S
.
and
Ong,
H.
Modif
ied fi
ref
l
y
al
gori
th
m
.
Journal
of
Ap
pli
ed
Mathe
m
ati
cs,
pp
.
1
-
12,
201
2.
[29]
W
ang,
G.,
Guo,
L.,
Duan
,
H.
,
L
iu,
L.
and
W
a
ng,
H.
,
A
Modi
fie
d
Fir
efly
Alg
orit
hm
for
UCA
V
Path
Planni
n
g.
Inte
rna
ti
ona
l
Jou
rna
l
of
H
y
brid
In
form
at
ion
T
ec
hn
olog
y
,
5(3)
,
pp
.
1
23
-
144,
2012
.
[30]
Kum
bhar
ana
,
S
.
and
Pand
e
y
,
G.
(2013).
S
olvi
ng
Tr
ave
l
ling
Sale
sm
an
Problem
using
Firefly
Algori
th
m
.
Int
ern
a
ti
ona
l
Jou
rna
l
for
Rese
arch i
n
Sc
ie
n
ce
& A
dvanc
ed Techn
ologi
es,
2(2), pp.53
--
57.
BIOGR
AP
H
I
ES
OF
A
UTH
ORS
Moath
Sababh
a
was
born
in
Irbi
d,
Jordan
in
198
7.
He
r
ec
e
ive
d
his
B.
Sc.
degr
e
e
in
El
e
ct
r
ic
a
l
Engi
ne
eri
ng
fro
m
Jordan
Univer
sit
y
of
Sc
.
and
Te
ch
.
JU
ST,
Irb
id,
Jordan
in
20
10.
He
recei
v
ed
his
M.
Sc.
degr
ee
in
El
e
ct
r
ic
a
l
Engi
nee
ring
fr
om
Oakla
nd
Un
ive
rsit
y
,
Ro
che
s
te
r,
Mic
h
iga
n,
U.S.A
in
2013
.
He
is
cur
r
ent
l
y
a
Ph.
D.
c
andi
d
at
e
in
El
e
ct
r
ic
a
l
and
Com
pute
r
Engi
ne
eri
ng
at
Oakla
nd
Univer
s
ity
,
Ro
che
ster
,
Michi
gan
,
U.S.
A.
His re
sea
rch
i
nte
rests
ar
e
in th
e
areas of
digital
signal
pro
ce
ss
in
g,
nonl
inear
est
i
m
at
ion
and
pre
d
i
ct
ion
,
fu
zzy
logic
and
de
ci
sion
m
aki
ng.
Moham
ed
A.
Zohd
y
w
as
born
in
Caro,
Eg
y
pt
.
He
r
ec
e
iv
ed
B.
Sc.
degr
e
e
in
E
le
c
trica
l
Engi
ne
eri
ng,
Ca
iro
Univer
sit
y
,
Eg
y
p
t
in
1968
and
he
rec
ei
v
ed
his
M.
Sc.
and
Ph
.
D.
degr
ee
s
in
El
e
ct
ri
ca
l
Engi
n
ee
ring
from
Univer
sit
y
o
f
W
at
er
loo,
Cana
d
a
in1
974,
and
1977
respe
ctively
.
He
worked
in
var
io
usindustrie
s:
do
wt
y
,
iron
and
s
te
e
l,
and
spar
.
He
is
cur
ren
tlya
Profess
or
at
Oakla
nd
Univer
sit
y
,
Roche
st
er
Hill
s,
Michi
g
an,
U.S.A.
His
rese
arc
h
intere
sts
are
in
ar
ea
s
ofc
o
ntrol, estima
ti
on,
comm
unic
a
ti
on,
neur
al
n
et
w
orks,
fuz
z
y
log
ic a
nd
h
y
br
id
s
y
ste
m
s.
Evaluation Warning : The document was created with Spire.PDF for Python.