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.
9
, No
.
5
,
Octo
ber
201
9
, pp.
3701
~3
713
IS
S
N: 20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v9
i
5
.
pp3701
-
37
13
3701
Journ
al h
om
e
page
:
http:
//
ia
es
core
.c
om/
journa
ls
/i
ndex.
ph
p/IJECE
Maximi
zin
g
sign
al
to
l
eakage r
atios
in MI
MO BC
H
cooperativ
e
beamf
ormin
g s
ch
eme
Moham
med
F
ad
hil
1
, N
or
F
adz
il
ah
A
bd
ull
ah
2
,
Mah
am
od Ism
ail
3
,
R
osdi
ad
ee
N
ordin
4
, C
e
brail
Ciftl
ikl
i
5
Musaab
Al
-
O
ba
idi
6
1
,2,3,4
Cent
re
of
A
dvanc
ed
Elec
tro
nic
and
Com
m
unic
a
ti
on
Engi
n
e
e
ring
(PA
KET)
,
Univer
siti
Keba
ngsaa
n
Mal
a
y
s
ia,
Mal
a
y
s
ia
5,6
Facul
t
y
of Eng
ine
er
ing, E
rc
i
y
es
Univer
sit
y
,
Tur
ke
y
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Ja
n
7
, 201
9
Re
vised
A
pr 12
, 2
01
9
Accepte
d
Apr 19
, 201
9
Bea
m
form
ing
(BF)
te
chni
qu
e
i
n
coope
ra
ti
ve
m
ult
iple
input
m
u
lt
iple
output
(MIM
O)
ant
enna
arr
a
y
s
improve
s
signal
to
noise
rat
io
(SN
R)
of
the
int
end
ed
user.
Th
e
cha
l
lenge
is
to
design
tra
nsm
it
be
amform
ing
vec
tors
fo
r
eve
r
y
user
while
li
m
iting
t
he
co
-
ch
anne
l
i
nte
rfe
r
ence
(CC
I)
from
othe
r
users.
In
thi
s
pape
r,
we
prop
osed
coope
r
at
iv
e
bea
m
form
ing
base
d
on
Signal
-
to
-
Leaka
g
e
Rat
io
(SLR)
to
expl
oit
the
leak
age
power
as
a
useful
power
in
the
se
cond
ti
m
e
slot
after
user
coope
ra
ti
o
n,
for
thi
s
purp
ose
succ
essive
int
erf
ere
n
ce
ca
nc
el
l
at
ion
(SI
C)
is
employ
e
d
in
ea
ch
user
to
sepa
rate
the
lea
kage
signal
from
the
desired
signal
.
W
it
ho
ut
inc
r
ea
sing
th
e
complexit
y
,
Maximizi
ng
Signal
-
to
-
L
ea
k
a
ge
Ra
ti
o
(SLR)
subjec
t
to
propo
sed
power
const
rai
nt
instead
of
a
unity
nor
m
is
the
wa
y
t
o
ac
hie
v
e
ex
tra
le
aka
g
e
power
.
To
re
du
ce
the
err
oneous,
Bose
–
Chaudhuri
–
Hocque
nghem
(BCH)
code
s
e
m
plo
y
ed
in
Bea
m
form
ing
of
(SIC)
coope
ra
ti
ve
sch
eme
BF
(CS
-
SIC
-
BCH).
Maximum
-
li
kelihood
(ML)
esti
m
at
or
m
et
h
od
is
used
at
each
user
recei
v
er.
Sim
ula
ti
o
n
result
s
show
tha
t
the
p
erf
orm
a
nce
of
the
prop
o
sed
sche
m
e
BF
(CS
-
SIC
-
BCH)
over
Ra
yle
igh
and
Ricia
n
fad
ing
ch
annel
is
signifi
c
antl
y
bet
t
er
tha
n
the
per
form
ance
bea
m
form
ing
base
d
on
SLR
in
Non
-
coope
rati
ve
s
y
stem
.
More
spec
if
i
call
y
to
a
chieve
a
BER
of
abou
t
10
−
4
the
req
u
ire
d
SN
R
for
the
proposed
sch
em
e
is
about
1
d
B
le
ss
th
an the
Non
-
coope
rative
s
y
st
em.
Ke
yw
or
d
s
:
Be
a
m
fo
rm
ing
Bose
–
C
ha
udhu
ri
–
Ho
c
quen
ghem
BC
H
Ca
ncell
at
ion
Chan
nel stat
e i
nfor
m
at
ion
Cooperati
ve
-
div
ersit
y
S
uccess
f
ul
-
inte
rf
e
ren
ce
Sign
al
-
to
-
le
a
ka
ge
-
rati
o
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
:
Moh
am
m
ed
Fadh
il
,
Ce
ntre of
Ad
va
nced El
ect
r
onic
an
d C
omm
un
ic
at
ion
En
gine
erin
g
(
PAKE
T),
Un
i
ver
sit
i Ke
ba
ngsaan
Mal
ay
sia
, 43600
U
K
M B
ang
i,
Sela
ngor, Mal
ay
sia
.
Em
a
il
:
m
oh
a
m
m
edf
ad
hilm
@
gm
ail.co
m
1.
INTROD
U
CTION
Mult
iple
anten
na
syst
e
m
s
i
mp
r
ove
the
sp
ec
tral
eff
ic
ie
ncy
without
increa
sing
po
wer
or
band
width,
theo
reti
cal
ly
,
t
he
capaci
ty
increases
with
th
e
nu
m
ber
of
a
nten
nas
de
plo
y
ed.
MIM
O
te
c
hn
i
qu
e
al
s
o
ga
ined
a
consi
der
a
ble
a
m
ou
nt
of
inte
r
e
st
as
te
ch
niqu
e
to
i
ncr
ease
th
e
data
rate
th
r
ough
sp
at
ia
l
m
ulti
plexing
or
e
nh
a
nce
the
qual
it
y
of
transm
issi
on
throu
gh
th
e
exp
l
oitat
ion
of
div
e
rsity
[1
]
.
In
m
ulti
us
er
MIM
O
dow
nlin
k
com
m
un
ic
at
ion
s
[
2]
,
a
ba
se
sta
ti
on
com
m
un
ic
at
es
with
se
ver
al
c
o
-
c
ha
nnel
us
er
s
i
n
the
sam
e
fr
equ
e
nc
y
and
tim
e
s
lots
.
T
he
refor
e
it
is
ne
cessary
to
de
sign
a
tra
ns
m
i
tt
er
that
able
to
sup
pr
ess
c
o
-
channel
inter
fe
ren
ce
(CCI)
because
co
-
c
ha
nn
el
int
erf
e
ren
ce
c
onsider
as
m
ajo
r
li
m
it
a
ti
on
facto
r
for
the
syst
em
capaci
ty
.
The
fo
c
us
of
this
pap
e
r
is
on
s
patia
l
dive
rsity
te
chn
iq
ue
s
in
a
dow
nlink
wireless
c
om
m
un
ic
at
ion
s
yst
e
m
,
wh
ere
a
base
sta
ti
on
(BS)
c
ould
sim
ultaneou
sly
serv
e
m
ulti
ple
us
ers,
wh
i
ch
re
qu
i
re
d
to
dep
l
oyed
beam
form
ing
[3
-
4]
at
BS
to
s
uppr
e
ss CC
I
to
end
us
e
rs
a
nd m
axi
m
iz
e overall
ca
p
aci
ty
.
Ma
ny
te
chn
iq
ue
s
hav
e
bee
n
adopted
t
o
sup
press
the
CC
I.
Fo
r
e
xam
ple
i
n
[
5],
the
pre
-
processi
ng
of
the
sig
nal
at
th
e
BS
was
em
plo
ye
d
to
com
plete
ly
cancel
the
CC
I
at
the
re
cei
ver
f
or
each
us
er
,
w
hile
in
[6
]
the
blo
c
k
-
diag
onal
iz
at
ion
was
pr
opos
e
d.
B
oth
[
5]
an
d
[
6]
rest
rict
ed
to
us
e
t
ran
sm
it
t
ing
a
nt
enn
as
t
o
be
great
er
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
20
19
:
3
7
0
1
-
3
7
1
3
3702
than
t
he
s
um
of
al
l
receivi
ng
a
nten
nas.
I
n
[
7]
the
s
pace
-
tim
e
blo
ck
c
odes
(S
TBC
)
ha
ve
bee
n
porpose
d
to
desig
n
the
pre
cod
e
r
as
a
wa
y
to
su
ppress
CC
I,
this
way
al
so
requires
a
la
rg
e
num
ber
of
a
nten
nas
at
BS
.
Othe
r
works
[8
-
11]
ha
ve
be
en
em
plo
ye
d
t
he
ze
ro
f
or
ci
ng
(ZF)
at
t
he
receiver
as
a
s
chem
es
fo
r
pe
rf
ect
ly
cancel
in
g
t
he
CC
I
f
or
e
ach
us
er
wh
ic
h
al
s
o
resit
ed
by
t
he
num
ber
of
transm
it
anten
nas
at
t
he
B
S.
So
m
e
researc
hers
sugg
e
st
ed
it
erati
ve
al
gorithm
s
to
so
lv
e
the
op
ti
m
iz
ation
prob
le
m
in
m
ul
ti
us
er
to
cancel
the
CC
I
[
12
-
13
]
.
Othe
r
sc
hem
es
based
on
opti
m
az
ti
on
te
ch
niques
to
m
axi
m
iz
e
the
outp
ut
sig
nal
-
to
-
interfe
re
nce
pl
us
-
noise
rati
o
(S
I
NR)
[
14
-
15
]
wh
ic
h
al
so
i
m
po
se
to
restr
ic
ti
on
on
t
he
nu
m
ber
of
t
ra
ns
m
it
anten
nas
t
o be
gr
eat
er
tha
n
t
he
num
ber
su
m
m
at
ion
of all
users’
an
te
nn
a
s.
Du
e
to
t
he
co
m
plexit
y
that
resu
lt
s
f
ro
m
c
oupled
natu
re
of
the
s
olv
i
ng
op
ti
m
iz
a
ti
on
pro
blem
of
m
axi
m
iz
ing
the
al
te
rn
at
ive
s
o
-
cal
le
d
sig
nal
-
to
-
i
nterf
e
re
nc
e
noise
rati
o
(
SI
NR
)
[16],
a
n
al
te
r
native
s
chem
e
base
d
on
the
con
ce
pt
of
sig
na
l
le
akag
e,
ha
ve
bee
n
sug
ge
ste
d
for
de
sig
ni
ng
tra
ns
m
it
bea
m
fo
rm
ing
ve
ct
or
s.
More
e
xpli
ci
tly
co
-
c
hannel
i
nterf
e
re
nce
(C
CI)
def
i
ne
as
t
he
inter
fer
e
nce
of
oth
e
r
us
er
s
on
the
desire
d
use
r,
wh
il
e the
lea
ka
ge defin
es as t
he
inte
rf
e
ren
ce
cause
d by the
sign
al
inte
nded
f
or a
desire
d u
ser on
t
he rem
a
inin
g
us
ers
.
Se
quent
ia
ll
y,
the
le
akag
e
is
a
m
easur
e
of
how
m
uch
sig
nal
po
wer
le
aks
into
t
he
oth
e
r
us
e
rs.
S
ever
al
stud
ie
s
base
d
on
m
axi
m
izing
th
e
sig
nal
-
to
-
le
aka
ge
-
a
nd
-
noise
r
at
io
(S
L
NR)
f
or
desi
gnin
g
t
he
b
eam
form
ing
vecto
r
sug
gest
ed
[
17
-
18]
w
hich
le
ad
s
to
a
decou
pled
op
ti
m
iz
a
ti
on
prob
le
m
and
adm
i
ts
an
analy
ti
cal
cl
os
ed
-
f
or
m
so
luti
on
.
M
or
e
ov
er,
the
s
olu
ti
on
of
the
le
a
kag
e
schem
e
do
es
not
restrict
by
th
e
nu
m
ber
of
B
s
an
d
us
ers
a
nten
nas
.
Anothe
r
sc
he
m
e
of
the
signa
l
to
le
akag
e
SLR
with
abse
nt
of
a
dd
it
ive
to
no
ise
powe
r
te
r
m
su
ggest
e
d by [
19
]
wh
ic
h
ai
m
s to desig
n bea
m
fo
r
m
ing
v
e
ct
or an
d
e
nhance
the
bit erro
r p
erfor
m
ance
.
Fo
ll
owin
g
the
[19],
we
pro
posed
beam
fo
rm
ing
base
d
o
n
si
gn
al
-
to
-
le
a
ka
ge
-
rati
o
(
SLR)
i
n
su
cce
ssive
interfe
ren
ce
ca
ncell
at
ion
(
SIC
)
co
op
e
rati
ve
schem
e.
W
e
a
i
m
to
exp
loit
the
le
aka
ge
power
as
a
us
e
f
ul
powe
r
instea
d
of
ig
no
rin
g
it
.
M
or
e
e
xp
li
ci
tl
y
we
use
SI
C
[
20
]
a
s
a
te
ch
nique
to
s
epar
at
e
t
he
des
ired
si
gnal
f
r
om
the
le
akag
e
si
gn
al
and
is
olate
the
le
akag
e
to
use
again
as
a
us
ef
ul
po
wer
f
ro
m
the
second
co
operati
ve
us
er
.
Sequentia
ll
y
w
e
update
t
he
powe
r
c
onstrai
nt
of
the
Ma
xim
al
Sign
al
-
to
-
Le
akag
e
Ra
ti
o
(SLR
)
prob
le
m
i
ns
te
ad
of
a
unit
y
nor
m
to
achieve
extra
le
aka
ge
powe
r.
T
o
re
duce
the
erron
e
ous,
Bo
se
–
C
haudhu
ri
–
Ho
c
que
nghem
(BCH)
c
odes
[21]
em
plo
ye
d
in
Be
a
m
fo
r
m
ing
of
(S
IC
)
co
op
erati
ve
sch
em
e
BF(CS
-
SI
C
-
BC
H).
Ma
xim
u
m
-
li
ke
li
ho
od
(ML)
es
tim
a
tor
m
et
ho
d
is
us
e
d
at
t
he
each
us
e
r
rece
iver.
Th
e
dr
a
w
back
of
t
he
pro
po
s
ed
schem
e
is
the
co
op
e
rati
on
wh
ic
h
m
eans
exch
a
nge
of
i
nfor
m
at
ion
.T
his
le
ads
to
the
requirem
ent
of
m
or
e
resou
rces
for
tr
ansm
issi
on
and
addit
ion
al
del
ay
s
.
Si
m
ulati
on
s
of
the
syst
em
are
carrie
d
out
ov
e
r
Ra
yl
ei
gh
(in
the
NL
O
S
en
vir
onm
ent
)
a
nd
Ri
ci
an
fad
i
ng
c
ha
nn
el
(in
th
e
LO
S
env
i
ronm
ent).
Re
su
lt
s
of
sim
ulati
on
dem
onstrat
e
that
the
perform
ance
of
the
b
eam
fo
rm
ing
base
d
on
SLR
in
SI
C
co
opera
ti
ve
schem
e
BF
(CS
-
SI
C
-
BC
H)
(
pro
posed
s
chem
e
)
ou
tpe
r
form
s
beam
fo
rm
ing
base
d o
n
SLR
of
Non
-
c
oope
rat
ive sc
hem
es
[19]
at
h
i
gh S
NRs.
The
rem
ai
nd
er
of
this
pa
pe
r
is
organ
iz
e
d
as
f
ollows.
I
n
sect
ion
2,
we
i
ntrod
uce
the
syst
em
m
od
el
for
dow
nlink
m
ulti
us
er
MIM
O
beam
fo
rm
ing
schem
es.
The
o
ptim
iz
a
ti
on
of
tra
ns
m
i
tt
i
ng
b
eam
fo
rm
i
ng
is
form
ulate
d
in
sect
ion
3
w
he
reas
Sig
nal
-
to
-
Leaka
ge
Ra
ti
o
(S
LR)
Ma
x
i
m
ining
is
de
s
cribe
d
in
sect
i
on
4.
All
these
t
he
n
us
ed
in
sub
-
sect
io
n
4.1
and
4.2
to
f
orm
ulate
the
pro
posed
sch
e
m
e
to
m
axi
m
iz
e
the
Sign
al
-
to
-
Leak
age
Ra
ti
o
(SLR
)
by
up
da
ti
ng
the
opti
m
iz
at
ion
of
transm
it
beam
form
ing
co
ns
t
raint.
The
desc
ribing
of
BC
H
co
din
g
te
ch
niques
a
re
pr
ese
nted
in
sect
ion
5,
wh
i
le
pro
po
se
d
do
wn
li
nk
c
oope
r
at
ive
schem
e
fo
r
tw
o
c
oope
rati
ve
us
ers
is
pr
ese
nt
ed
in
s
ect
ion
6.
In
sect
io
n
7,
the
c
hannel
m
od
el
f
or
t
he
fir
st
an
d
seco
nd
ti
m
e
sl
ot
is
descr
ibe
d.
T
he
res
ults
are
us
e
d
to
co
m
par
e
the
BE
R
per
f
or
m
ance
of
the
beam
fo
rm
ing
base
d
on
SL
R
in
SI
C
c
oope
rati
ve
sc
he
m
e
BF(CS
-
S
I
C
-
BC
H)
w
it
h
beam
fo
rm
ing
base
d
on
S
LR
of
Non
-
co
operati
ve
syst
em
are
dr
a
w
n
in
secti
on
8.
2.
MU
LT
I
US
E
R
M
I
MO
BE
A
MFO
R
MING
In
dow
nlin
k
MU
-
MIM
O
be
a
m
fo
rm
ing
[
19
]
,
a
bas
e
s
ta
ti
on
(BS
)
e
q
ui
pp
e
d
with
M
anten
nas
com
m
un
ic
at
es
with
U
M
ulti
-
anten
na
us
e
rs.
Each
us
e
r
rece
ived
an
d
tran
s
m
it
the
signa
l
ind
e
pende
ntly
us
in
g
U
anten
na,
the
total
us
ers
’
an
te
nn
as
=
∑
=
1
as
sh
ow
n
in
Fig
ur
e
1.
I
n
a
wire
le
ss
com
m
un
i
cat
io
n
netw
ork,
t
he
t
ypic
al
syst
e
m
assum
es
that
≥
in
in
dep
e
nde
nt
cha
nnel
s
of
flat
fa
ding.
T
he
inten
de
d
data
sign
al
for use
r u is t
he
scal
ar
, so t
he
tra
ns
m
itted sym
bo
l vec
tor
t
o U
us
ers
is:
=
[
1
,
2
,
…
,
]
(1)
Wh
il
e th
e
pr
ec
od
i
ng m
at
rix
is de
fine
d
as,
=
[
1
,
2
,
…
,
]
(2)
Wh
e
re
∈
C
x1
is t
he
b
eam
fo
rm
ing vecto
r for
u
th
us
er
.
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
Maximizi
ng
sig
na
l t
o
le
ak
age
ra
ti
os
i
n
M
IM
O
BC
H
c
ooper
ative
b
e
am
f
ormin
g
sc
he
me
(
Mo
hamm
e
d Fa
dh
il
)
3703
Figure
1. Mult
i
-
us
e
r beam
fo
r
m
ing
syst
em
The
tra
ns
m
it
ted
vector
will
r
esult
from
m
ulti
plyi
ng
,
the
beam
fo
rm
ing
vecto
r
by
t
he
d
at
a
sym
bo
l
would be:
=
∑
=
=
1
(3)
The
tra
ns
m
itte
d
sig
nals
WS
∈
x1
are
broadc
ast
ing
ov
e
r
al
l
channels
between
B
S
an
d
us
er
s
denoted
as:
=
[
1
,
2
,
…
,
]
(4)
w
he
re
∈
x
is
the
channel
c
oeffici
ents
betw
ee
n
received
a
nt
enn
a
s
of
u
t
h
use
r
an
d
M
a
ntenn
a
s
of
BS as:
=
[
ℎ
(
1
,
1
)
…
ℎ
(
1
,
)
⋮
⋱
⋮
ℎ
(
,
1
)
…
ℎ
(
,
)
]
(5)
w
he
re
ℎ
(
,
)
represen
t
s
the
cha
nn
el
coeffic
ie
nt
that
eff
ect
on
t
he
pro
pag
at
io
n
si
gn
al
betwee
n
m
th
tra
ns
m
it
te
r
arr
ay
a
nten
na
of
B
S
an
d
n
t
h
recei
ver
ar
r
ay
anten
na
of
the
u
t
h
us
er
.
Th
us
,
t
he
rec
ei
ved
si
gnal
s
by
the
receiver
s’
a
nte
nn
a
s:
=
[
1
,
2
,
…
,
]
=
+
(6)
By
co
ns
ide
rin
g
∈
x
1
as
t
he
sig
na
l,
w
hic
h
is
re
c
ei
ved
at
t
he
i
th
reci
pient,
w
hil
st
f
or
the
a
ddit
ive
no
ise
is
denot
ed
by
n
∈
x
1
.
When
we
ha
ve
c
onside
red,
eac
h
us
e
r
se
pa
ratel
y,
we
will
fi
nd
th
e
re
cei
v
e
d
sign
al
at a
n
i
th
recipient a
s:
=
∑
+
=
1
=
+
∑
+
=
1
,
≠
(
7
)
The
vect
or
ha
s
com
plex
Ga
us
sia
n
va
riable
s
Com
po
ne
nts
with
(unit
–
var
i
ance)
an
d
(ze
r
o
-
me
an)
.
More
ov
e
r,
t
he
com
po
ne
nts
of
the
ad
diti
ve
noise
hav
e
distr
ibu
ti
on
as
(0,
σ
2
)
and
is
te
m
po
ra
rily
wh
it
e
an
d
sp
at
ia
ll
y.
To
de
scribe
t
he
pro
po
s
ed
sc
hem
e
cl
early
,
as
fo
ll
ow
i
ng
we
rev
i
ew
the
or
i
gin
al
SLR
base
d
precod
i
ng
schem
e
[19]
i
n sec
ti
on 4.0.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
20
19
:
3
7
0
1
-
3
7
1
3
3704
3.
TRA
NSMIT
B
EAMF
ORM
ING
OPTI
MI
Z
ATION
On
e
of
t
he
m
a
in
ste
ps
to
e
nhance
syst
em
p
erfor
m
ance
is
transm
itti
ng
be
a
m
fo
rm
ing
optim
iz
at
ion
.
Accor
ding
to
[22]
the
m
axi
m
iz
ing
of
s
om
e
arb
it
rar
y
ut
il
i
ty
fu
nctio
n
(
SINR
1
,
.
,
SINR
U
)
can
stric
tl
y
be
incr
easi
ng
in
the
SINR
of
each
us
e
r,
a
nd
the
total
t
ra
ns
m
it
po
we
r
is
r
est
rict
ed
by
.
Ma
them
at
icall
y
sp
ea
king is,
ma
x
(
SINR
1
,
.
.
.
,
SINR
U
)
subj
ect
to
∑
‖
w
u
‖
2
≤
U
u
=
1
(8)
4.
MA
X
I
MIZ
IN
G SIGN
AL T
O LEAK
AGE
RATIO
In
this
wor
k
,
t
he
si
ng
le
us
e
r
m
axi
m
u
m
-
l
ikeli
ho
od
SINR
is
co
ns
ide
red,
as
sho
wn
in
e
qu
at
io
n
(
9).
Using
SINR
in
(
9)
f
or
=
{
1
.
.
.
.
}
as
a
n
op
ti
m
iz
ation
obj
ect
iv
e
f
unct
ion
f
or
determ
i
ning
t
he
{
}
=
1
will
lead a
pro
blem
w
it
h
U
c
oupled
v
a
riabl
es {
}
[19]
.
SINR
=
|
|
|
|
2
2
+
∑
|
|
∗
̃
∗
̃
|
|
2
=
1
,
≠
|
|
|
|
2
(
9)
Fo
r
the
a
bove
reason
in
design
of
the
b
ea
m
fo
r
m
ing
coe
f
fici
ents
{
w
i
}
SL
R
are
s
ugge
ste
d
in
[
19
]
,
wh
ic
h
re
s
ult
in
a
fu
ll
char
act
erizat
ion
of
th
e
op
ti
m
al
so
lut
ion
s
in
te
rm
s
of
ge
ner
al
iz
ed
e
igen
value
pro
bl
e
m
s.
Wh
e
re
SLR
is
the
rati
o
of
t
he
powe
r
of
the
desire
d
si
gn
al
|
|
H
i
w
i
|
|
2
to
t
he
powe
r
of
the
i
nterf
e
re
nce
ca
us
e
d
by this
us
e
r
i
on the
sig
nal re
c
ei
v
ed by
us
er
u,
|
|
H
u
w
i
|
|
2
.
=
|
|
|
|
2
∑
|
|
|
|
2
=
1
,
≠
(10)
4.1
.
Pr
ob
le
m
statemen
t
(P
1):
By
m
axi
m
izing
S
LR t
o
c
om
pu
te
m
axi
m
u
m
beam
fo
rm
ing
(
w
i
o
)
f
or eac
h user
accor
ding t
o
[
19]
.
=
|
|
|
|
2
∑
|
|
|
|
2
=
1
,
≠
(
P1
)
subj
ect
to
‖
w
‖
2
=
1
=
{
1
,
.
.
.
}
w
he
re
|
|
|
|
2
re
pr
e
sen
ts
the
re
quire
d
sign
al
po
wer
of
us
e
r
i
,
w
hile
∑
|
|
|
|
2
=
1
.
≠
re
pr
esents
the
tota
l
le
akag
e
powe
r
from
the
total
powe
r
of
us
e
r
i
as
interfer
e
nc
e
on
t
he
ot
her
us
ers
.
By
caref
ully
loo
ki
ng
t
o
SLR
in
(P1).
It’s
ea
sy
to
s
ay
t
hat
for
,
=
{
1
.
.
.
.
}
the
U
is
de
coupled
opti
m
iz
at
ion
pro
bl
e
m
s
com
par
in
g
with e
qu
at
i
on (9).
4.2
.
Pr
oposed
scheme
(P
2):
Fo
ll
ow
the
ge
ner
al
opti
m
iz
at
ion
of
transm
it
bea
m
fo
rm
ing
co
ns
trai
nt
in
equ
at
io
n
(
8)
and
[2
3
]
,
we
update
t
he
SL
R
co
ns
trai
n
i
n
(P1
)
t
o
be;
‖
‖
2
=
≤
/
.
Where
/
is
tra
ns
m
issi
on
po
we
r
const
raint
at
the
tra
ns
m
it
te
r
,
w
hic
h
ca
n
be
desc
ribe
d
a
s
(
‖
‖
2
)
≤
.
The
sym
bo
l
s
i
sat
isfie
s
th
e
powe
r
Co
ns
trai
nt as
=
(
|
|
2
)
=
1
.
The
reason
for
this
up
dating
is
t
he
draw
bac
k
of
t
he
c
onstrai
nt
in
pro
ble
m
s
ta
tem
ent
(P
1)
is
wh
e
n
each
us
er
has
m
ul
ti
ple
data
s
tream
s,
the
effe
ct
ive
cha
nnel
gain
f
or
eac
h
s
tream
can
be
s
ever
el
y
unbala
nced.
If
powe
r
co
ntr
ol
or
a
da
ptive
m
odulati
on
an
d
cod
i
ng
can
not
be
a
pp
li
ed
,
the
overall
e
r
ror
perform
ance
of
eac
h
us
er
w
il
l s
uffer si
gnific
ant l
oss
[2
4
]
.
S
o by updatin
g
t
he
c
onstrai
nt,
of
(P2
),
we get
(P2
)
a
s foll
ow
i
ng
:
=
|
|
|
|
2
∑
|
|
|
|
2
=
1
,
≠
(P
2)
Subject
to
‖
‖
2
=
≤
/
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
Maximizi
ng
sig
na
l t
o
le
ak
age
ra
ti
os
i
n
M
IM
O
BC
H
c
ooper
ative
b
e
am
f
ormin
g
sc
he
me
(
Mo
hamm
e
d Fa
dh
il
)
3705
It
is
note
d
that
the
norm
of
is
ir
releva
nt
t
o
the
final
so
l
ution
s
,
or
i
n
ot
he
r
w
ords,
the
norm
of
can
be
force
d
to
be
any
val
ue
to
ac
hieve
th
e
best
value
f
or
unde
r
the
powe
r
c
onstrai
nt
.
By
substi
tuti
ng
̃
=
∑
=
1
,
≠
into
obj
ect
ive
fun
ct
io
n o
f
(
P2)
, w
e
can
obtai
n;
=
|
|
|
|
2
|
|
̃
|
|
2
=
∗
∗
∗
̃
∗
̃
(11)
Our
up
dating
dep
e
nds
on
tha
t
the
gen
e
ral
s
olu
ti
on
of
(
P2)
wh
ic
h
is
so
lv
ed
by
[
19]
is
obey
ed
to
t
he
Ra
yl
ei
gh
–
Ri
tz
m
et
ho
d
[25]
wh
ic
h
sat
isfy
ing
ly
any
co
nst
raint
without
eff
ect
on
the
gen
e
ral
so
luti
on.
S
o,
fo
ll
owin
g [19]
the s
olu
ti
on
of
(11) will
b
e:
∗
∗
∗
̃
∗
̃
≤
(
∗
,
̃
∗
̃
)
(12)
w
he
re
λ
ma
x
is
the
la
rg
est
ge
ner
al
i
zed
ei
ge
nval
ue
.
Acc
ordi
ng
to
the
SLR
crit
eri
on,
the
prec
od
i
ng
m
at
rix
is
desig
ne
d based
on th
e
foll
owing m
et
ric;
∝
.
(
∗
,
̃
∗
̃
)
(13)
At
us
e
r
i
,
m
ax
im
u
m
-
li
kelih
ood
(ML)
det
ect
ion
sc
hem
e
will
be
us
ed
to
est
i
m
at
e
(
s
)
from
the
receive
d
sig
nal
as foll
owin
g
[
19
]
̃
=
∗
∗
|
|
|
|
2
(14)
t
hen
̃
=
+
∗
∗
∑
=
1
,
≠
|
|
|
|
2
+
∗
∗
|
|
|
|
2
(15)
5.
BCH
CODI
N
G TECH
NI
Q
UE
The
BC
H
e
nc
od
e
r
blo
c
k
cre
at
es
a
BC
H
cod
e
with
m
essage
le
ngth
=5
and
c
odew
ord
le
ng
t
h
=
15.
T
he
in
p
ut
m
us
t
con
ta
in
exactl
y
el
e
m
ents.
The
outp
ut
is
a
vecto
r
of
l
eng
t
h
m
us
t
ha
ve
the
f
or
m
2^(
(
-
1))
,
wh
e
re
is
an
i
nteger
gr
eat
e
r
tha
n
or
e
qual
to
3.
For
a
giv
e
n
code
w
or
d
le
ng
th
, only s
pe
ci
fic m
essage
le
ng
th
s
are
val
id for a BC
H
c
od
e
.
The
BC
H
dec
od
e
r
bl
oc
k
rec
ov
e
rs
a
bin
ary
m
essage
vect
or
from
a
bin
ary
BC
H
co
de
word
vect
or
.
Fo
r
pro
per
de
cod
i
ng,
the
fi
r
st
two
pa
ram
e
te
r
values
i
n
this
blo
c
k
s
hould
m
at
ch
the
par
am
et
ers
i
n
th
e
corres
pondin
g
BC
H
encode
r
blo
c
k.
T
he
input
is
a
bina
ry
cod
e
w
ord
vector
a
nd
t
he
first
outp
ut
is
th
e
corres
pondin
g
bin
a
ry
m
essage
vect
or
.
I
f
t
he
BC
H
co
de
has
m
essage
le
ngt
h
an
d
c
odew
ord
le
ngth
,
then
the
in
put
has
le
ngth
a
nd
t
he
first
out
put
has
le
ngth
.
I
f
t
he
in
pu
t
is
fr
am
e
-
based,
then
it
m
us
t
be
a
colum
n
vector
.
T
he
sec
ond
outp
ut
is
the
num
ber
of
error
s
detect
ed
du
ri
ng
dec
od
i
ng
of
the
c
od
e
word.
A
neg
at
ive
int
eger
in
dicat
es
that
the
bl
oc
k
detect
ed
m
or
e
er
r
or
s
tha
n
it
cou
l
d
c
orr
ect
us
in
g
t
he
cod
i
ng
schem
e. Th
e sa
m
ple t
i
m
es o
f a
ll
inp
ut
an
d o
utput sig
nals a
r
e eq
ual.
6.
SY
STE
M MO
DEL
In
our
pr
opos
e
d
dow
nlink
coo
pe
rati
ve
syst
e
m
s
m
od
el
,
two
co
op
e
rati
ve
use
rs
U
=2
,
with
N
ante
nna
arr
ay
f
or
eac
h
us
er
c
omm
un
ic
at
e
with
M
anten
na
base
st
at
ion
(BS
).
T
he
cooper
at
ive
schem
e
inv
olve
s
two
ste
ps
as
sho
wn
in
Fi
g
ure
2.
I
n
the
fir
st
tim
e
slot,
t
he
data
is
dec
oded
on
the
tra
ns
m
it
ter
side
,
wh
e
re
bin
a
ry
seq
uen
ces
-
bit
s
lo
ng
a
re
ge
ne
rated
by
a
pse
udo
-
ra
ndom
gen
e
rato
r
bl
oc
k
w
hich
is
the
input
to
t
he
B
CH
encode
r
bl
ock.
The
e
ncode
r
bl
ock
m
aps
-
bit
s
of
t
he
se
qu
e
nc
e
into
bits
of
the
seq
ue
nce.
The
represe
nt
the
da
ta
sig
nal
f
or
us
er
u
is
the
.
The
data
sig
na
l
a
re
passe
d
on
t
he
QPSK
m
odulato
r,
the
n
eac
h
u
ser
dat
a
m
o
dul
a
ted
per
form
bea
mfor
mi
ng
as
,
T
he
Gau
s
sia
n
-
li
ke
(
A
WGN
an
d
Ra
yl
ei
gh
)
c
ha
nn
el
bl
ock
is
desig
ne
d
to
int
rod
uce
a
fa
ding
e
ff
ect
a
nd
a
dd
n
oise
t
o
t
he
m
od
ula
te
d
sign
al
as
sh
ow
n
in
(
7).
In
the
sec
ond
tim
e
slot,
e
ach
us
e
r
retra
ns
m
itted
the
data
to
his
par
t
ner
[2
6
]
.
D
ue
to
the
com
plexity
of
hard
war
e
im
ple
m
entat
ion
whic
h
use
d
f
or
li
near
detect
io
n,
we
sug
gested
the
S
IC
m
et
ho
d
for
i
m
pr
ovin
g
the
det
ect
ion pe
rfo
rm
ance w
it
h
lo
w
c
om
plexity
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
20
19
:
3
7
0
1
-
3
7
1
3
3706
Figure
2.
Be
am
fo
r
m
ing
coo
per
at
ive
sche
m
e
betwee
n
t
wo
us
er
s
em
plo
yi
ng
S
IC si
gn
al
de
te
ct
ion
a
nd ML
est
i
m
ation
The
basic
idea
of
S
IC
[
20]
is
t
hat
us
er
sig
nal
s
are
s
uccessiv
el
y
decode
d
by
ML
detect
or
to
est
im
at
e
s
i
from
the
received
sig
nal
.
A
f
te
r
one
use
r
’s
sign
al
is
decoded,
it
is
subtr
act
ed
f
ro
m
the
com
bin
ed
si
gnal
befor
e
t
he
ne
xt
us
er
’s
sig
nal
is
decode
d.
When
SI
C
is
ap
pli
ed,
on
e
of
t
he
us
er
si
gn
al
s
is
decode
d,
treat
i
ng
t
he
oth
e
r
use
r
si
gnal
as
an
i
nterf
e
rer
,
but
the
la
t
te
r
is
the
n
dec
od
e
d
with
t
he
ben
e
fit
of
the
sign
al
of
t
he
f
or
m
er
hav
i
ng alrea
dy b
ee
n rem
ov
ed.
Ma
them
atical
l
y spea
king the
rem
ai
nin
g
sig
na
l
,
is,
̃
=
−
.
More
ex
plici
tly
at
the
first
ti
m
e
slot,
ML
e
s
tim
a
ti
on
a
nd
S
IC
de
te
ct
ion
is
us
e
d,
w
he
re
t
he
(S
IC
)
will
be
us
e
d
to
dete
ct
the
inte
rf
e
re
nce
sym
bo
ls,
wh
e
re
eac
h
use
r
rec
og
nizes
a
nd
ide
ntifie
s
it
s
ow
n
sig
nal
(
d
esi
re
d
sign
al
)
from
ot
her
us
er
sig
nal
s
wh
ic
h
are
kn
own
as
the
le
akag
e
si
gn
al
,
th
en
the
le
aka
ge
sign
al
s
f
or
fi
r
st
and
seco
nd
us
e
rs
a
re
tra
ns
m
itted
to
his
pa
rtne
r
at
the
sec
ond
ti
m
e
slot.
By
us
i
ng
propose
d
sc
hem
e,
we
ca
n
e
xp
l
oit
the
po
wer
of
le
akag
e
sig
nal
t
o
get
m
or
e
pr
oductive
powe
r
by
com
bin
ing
t
he
desire
d
si
gn
a
l
(which
is
det
ect
ed
by
each
use
r
)
with
le
aka
ge
s
ign
al
(
wh
ic
h
de
te
ct
ed
by
the
seco
nd
us
er
).
I
n
ot
her
word
s
,
com
bin
ing
th
e
own
us
er
sig
nal
with
the
inter
fer
e
nc
e
sig
nal co
m
i
ng fro
m
h
is
part
ner
as
s
how
n
i
n
Fi
gure
2.
Wh
e
n
the
inte
r
-
us
e
r
cha
nnel
(
the
cha
nn
el
be
tween
us
ers
)
is
ta
ken
into
acc
ount,
m
or
e
po
wer
will
be
al
locat
ed
f
or
c
oope
rati
on.
I
n
si
m
ple
word
s
,
to
ta
ke
a
dvant
age
of
t
he
le
ak
ed
sig
nal,
we
hav
e
rei
ntrod
uc
ed
th
e
sign
al
t
hat lea
ke
d from
the o
ri
gin
al
si
gn
al
t
o i
ts real desti
nat
ion
.
Ther
e
a
re
tot
al
transm
issi
o
n
po
wer
c
on
strai
nts
at
th
e
transm
it
te
r,
wh
ic
h
can
be
de
scribe
d
as
(
‖
‖
2
)
≤
,
is a co
ns
ta
nt to
m
eet
the total t
ran
sm
it
ted
powe
r
c
on
st
r
ai
nt and it i
s
giv
en
as
[26]
:
=
√
(
−
1
(
−
1
)
)
(16)
Gen
e
rall
y,
in
m
ul
ti
us
er
c
ooper
at
ive
,
a
nd
a
ccordin
g
to
[
26]
.
̂
=
.
Wh
ere
t
he
receive
d
sy
m
bo
ls
are
pr
ece
de
d
w
it
h
pre
-
e
qu
al
iz
at
ion
weig
ht
,
Wh
e
re
=
−
1
.
The
tra
ns
m
it
ted
si
gn
al
t
o
u
th
us
er
at sec
ond t
i
m
e slot is:
̂
−
−
2
=
−
−
1
(17)
w
he
re
−
−
1
is
the
le
akag
e
sig
nal
fr
om
i
th
us
er
wh
ic
h
detect
e
d
by
uth
use
r
at
the
first
ti
m
e
slot
.
The
receive
d
sig
nal
at the sec
ond t
i
m
e slot in ith
us
er
is
giv
e
n b
y:
−
−
2
=
−
−
2
̂
−
−
2
+
(18)
w
he
re
−
−
2
is
the
inter
-
us
e
r
c
ha
nnel
betwe
e
n
ut
h
us
e
r
a
nd
the
it
h
us
e
r
an
d
is
the
A
WGN
in
it
h
use
r.
In
it
h
us
e
r,
m
a
xim
u
m
rati
o
com
bin
er
(MRC
)
will
be
us
ed
t
o
c
om
bin
e
the
desire
d
sig
na
l
̂
(
w
hich
dete
ct
ed
by it
s self as
it
s own sig
nal at
f
irst t
im
e slot) w
it
h
̂
−
−
2
.
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
Maximizi
ng
sig
na
l t
o
le
ak
age
ra
ti
os
i
n
M
IM
O
BC
H
c
ooper
ative
b
e
am
f
ormin
g
sc
he
me
(
Mo
hamm
e
d Fa
dh
il
)
3707
By
us
ing
t
he
MR
C
schem
e
after
fir
st
an
d
seco
nd
-
tim
e
slot
an
d
em
pl
oyed
Ma
xim
u
m
-
l
ikeli
ho
od
(ML)
e
stim
at
or
, th
e
sig
nal of t
he
it
h use
r wil
l be:
=
̂
+
∑
̂
−
−
2
=
1
,
≠
=
̂
+
∑
−
−
2
∗
|
|
−
−
2
|
|
2
−
−
2
=
1
,
≠
(19)
Finall
y,
the
re
cei
ved
si
gnal
f
or
eac
h
us
e
r
is
dem
od
ulate
d
by
the
QP
S
K
dem
od
ulato
r.
The
n
the
dem
odulate
d
sign
al
s a
re
pas
sed
t
o
BC
H
d
e
cod
e
r bloc
k
t
o reco
ver the t
ra
ns
m
itted signal
.
7.
CHAN
NEL M
ODEL
Du
e
to
LO
S p
r
o
pa
gatio
n
,
the strong
e
st
pro
pa
gation
c
om
po
nen
t o
f
MIM
O
cha
nnel
co
rr
e
s
ponds
t
o
th
e
determ
inist
ic
com
po
ne
nt
(also
refe
r
red
to
as
sp
ecular
com
po
ne
nts).
On
the
oth
e
r
hand,
al
l
the
oth
er
com
po
ne
nts
a
re
ra
ndom
com
po
nen
ts
(du
e
to
NL
OS
a
lso
re
ferred
t
o
as
scat
te
ri
n
g
com
pone
nts)
[2
7
].
The
broa
dcast
channel
distri
buti
on
has
bee
n
fo
ll
owin
g
the
Ra
yl
ei
gh
channel
distrib
utio
n
w
hich
is
Ga
us
sia
n
distrib
ution
w
it
h
a
var
ia
nc
e
of
σ
2
and
zer
o
m
ean.
T
ha
t
m
eans
ther
e
is
no
com
pone
nt
of
L
OS
(
R
ic
ia
n
=
0)
:
σ
=
√
1
Ric
ian
+
1
.
O
n
t
he
oth
e
r
ha
nd,
wh
e
n
the
re
is
any
com
ponent
of
L
OS
(For
R
ic
ia
n
>
0)
the
broa
dcast
c
hannel
distri
buti
on
has
been
f
ollow
i
ng
t
he
G
aussian
distrib
u
ti
on
with
a
va
riance
of
σ
2
a
nd
m
ean
of
q
or
Ri
ci
an
distri
bu
ti
on
wh
e
n
R
ic
ia
n
increase
s
as:
q
=
√
Ric
ian
Ric
ian
+
1
,
σ
=
√
1
Ric
ian
+
1
.
Ther
e
f
or
e,
in this
w
ork,
t
he
cha
nn
el
m
at
rix
of
th
e
MIM
O
syst
em
tend
s
to be
descr
i
be
d
as
[
27
]
:
H
=
√
R
i
c
ia
n
R
ic
ia
n
+
1
H
d
+
√
1
R
ic
ia
n
+
1
H
r
(20)
w
he
re
H
d
re
pr
ese
nting
the
c
om
po
nen
t
of
the
norm
al
iz
e
d
determ
inist
i
c
cha
nnel
m
at
rix,
w
hile
H
r
represe
nting
the
c
om
po
ne
nt
of
ran
dom
channel
m
at
rix,
with;
|
|
H
d
|
|
2
=
N
T
M
,
|
|
E{
|
[
H
r
]
i
,
j
|
2
}=
1,
i
=
1:
N
T
,
j
=
1:
M
[2
8
]
,
w
hile
R
i
c
ia
n
is
kn
own
as
fact
or
of
the
Ri
ci
an
c
hannel
w
hich
i
s
the
relat
ion
be
tween
the co
m
pone
nt of th
e
sp
ec
ular
pow
e
r
c
2
an
d
t
he
co
m
p
onent
of scatt
erin
g po
wer
2σ
2
, d
is
play
ed
as:
R
ic
i
a
n
=
‖
H
d
‖
2
E
{
|
[
H
r
]
i
,
j
|
2
}
=
c
2
2
σ
2
(21)
8.
SIMULATI
O
N RESULTS
The
pro
pose
d
syst
e
m
pr
esented
in
sect
ion
6
is
si
m
ula
te
d
us
ing
Ma
tl
ab
co
des.
Wh
e
re
the
down
li
nk
transm
itted sign
al
from
BS is
receive
d
by two
u
se
rs
as r
ecei
ver
s
. One
of
th
ese us
ers
will
act as a receiver
us
e
r
wh
il
e
the
oth
e
r
us
er
will
act
as
a
relay
us
er
and
vice
ve
rsa.
Ther
e
f
or
e,
t
he
re
are
tw
o
c
ha
nn
el
s
,
first
-
tim
e
slot
channel
bet
we
en
BS
a
nd
t
he
receiver
s
(
down
li
nk
c
h
an
ne
l)
an
d
seco
nd
-
tim
e
slot
chan
nel
betwe
en
t
he
us
er
s
(inter
-
us
e
r
c
ha
nn
el
s
).
T
he
dow
nlin
k
c
hannel
is
si
m
ula
te
d
as
R
ay
le
igh
cha
nn
el
with
ze
ro
-
m
ean,
wh
il
e inter
-
use
r
cha
nn
el
s sim
ulate
d
as Ri
ci
an
fa
ding ch
a
nn
el
w
it
h
m
-
m
ea
n
an
d
unit
-
var
i
ance in
dep
e
nd
ent and
id
entic
al
ly
d
ist
rib
uted (i.i
.d) c
om
plex
Ga
us
si
an
ra
nd
om
v
ariables.
QP
S
K
si
gn
al
c
onste
ll
at
i
on h
as
b
ee
n us
ed
a
s
a
broa
dcast
m
od
ulati
on
in
al
l
si
m
ulati
on
s
an
d
the
res
ults
ar
e
ave
rag
e
d
t
hroug
h
se
ver
al
c
hannel
in
vestig
at
ions
.
Fo
r
al
l
recei
ve
rs,
the
noise
var
ia
nc
e
pe
r
receive
a
nten
na
is
su
ppose
d
the
eq
ual,
1
2
=
.
.
.
=
2
=
2
.
The
s
umm
ary
of p
a
ram
et
ers
is sh
own
in Ta
bl
e 1
.
The
BER
perf
or
m
ance
of
al
l
the
syst
em
s
descr
ibe
d
is
e
valuated
at
BER
10
−
4
.
A
n
acce
pta
ble
BER
perform
ance
fo
r
vo
ic
e
c
omm
un
ic
at
io
n
is
10
−
4
[19]
,
[2
4
]
.
T
he
BER
perfo
rm
a
nce
of
the
bea
m
fo
r
m
ing
base
d
on
S
LR
in
SIC
cooper
at
ive
schem
e
BF(
CS
-
SI
C
-
BC
H
)
(
pro
pose
d
sch
e
m
e),
com
par
ing
with
beam
form
ing
base
d
on
SLR
in
Un
c
oded
-
N
on
c
oope
rati
ve
syst
e
m
[19]
as
sh
ow
n
in
Fig
ur
e
3.
T
he
ab
breviat
io
ns
us
e
d
in
the
si
m
ulati
on
r
es
ul
ts are li
ste
d
in
Table
2
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
20
19
:
3
7
0
1
-
3
7
1
3
3708
Table
1.
Sim
ul
at
ion
par
am
et
e
rs
Para
m
eters
Def
in
itio
n
Mod
u
latio
n
m
o
d
e
QPSK
No
.
o
f
inp
u
t data
1
0
0
0
0
BER co
m
p
ass
io
n
Poin
t
10
−
4
Do
wn
lin
k
chan
n
el
Rayleigh
Inter
-
u
ser
chan
n
els
Rician
SNR o
f
inter us
er
ch
an
n
el
5
–
2
5
Nu
m
b
e
r
o
f
us
ers (
U
)
2
Nu
m
b
e
r
o
f
anten
n
as f
o
r
BS (
M
)
2
,4,6
Nu
m
b
e
r
o
f
anten
n
as f
o
r
each u
se (N
)
2
,4,5
,6
Rician
chan
n
el f
acto
r
(
K
R
ici
an
)
10
–
25
Beta
(
β
)
Bo
se
–
Ch
au
d
h
u
ri
–
Ho
cq
u
en
g
h
e
m
(
B
CH)
0
.1
–
0
.9
[
1
5
,5]
Table
2.
Sim
ul
at
ion
a
bbre
viati
on
s
Para
m
eters
Def
in
itio
n
Bea
m
f
o
r
m
in
g
of
SI
C in
Unco
d
ed
coo
p
erative sch
e
m
e
BF(CS
-
S
IC)
Bea
m
f
o
r
m
in
g
of
SI
C in
Co
d
ed
co
o
p
erative sch
e
m
e.
BF(CS
-
S
IC
-
BC
H)
No
.
o
f
trans
m
itted
an
ten
n
as
T
No
.
o
f
r
eceived
ant
en
n
as
R
No
n
-
co
o
p
erative bea
m
f
o
r
m
in
g
BF
SNR o
f
inter
-
u
ser
ch
an
n
els d
B
SNRd
Bin
Facto
r
o
f
the Ricia
n
chan
n
el
Kd
B
Compari
son wi
th o
t
her
re
sear
ch
Figure
3
s
how
s
the
bit
error
r
at
io
(BER)
pe
r
form
ance
for
t
he
dow
nl
ink
of
the
pr
opos
e
d
schem
e
and
Non
-
co
operati
ve
syst
em
in
case
Bs
anten
na
assigne
d
with
M
=
4
an
d
eac
h
use
r
assi
gn
e
d
with
N=
5,
i
n
Ri
ci
an
inter
-
us
er
c
hannel
S
NR=
20
with
R
ic
ia
n
=
25
a
nd
β
=
0.1
.
Wher
e
Fig
ure
3,
de
m
on
strat
e
the
perform
ance
of
beam
fo
rm
ing
base
d
on
SLR
of
S
IC
U
nc
oded
-
c
ooper
at
i
ve
schem
e
BF(CS
-
SI
C
)
is
be
tt
er
than
U
nc
od
e
d
-
Nonc
oope
rati
ve
syst
e
m
[19]
.
More
s
pecifica
ll
y,
to
achieve
a
BER
of
a
bout
10
−
4
the
re
quir
ed
S
NR
f
or
t
he
BF(CS
-
SI
C)
i
s
about
3
dB
le
ss
than
the
Un
c
od
e
d
-
N
onco
operati
ve
s
yst
e
m
[19]
.
On
oth
e
r
ha
nds,
the
perform
ance
of
beam
fo
rm
ing
of
based
on
S
LR
in
S
IC
c
oo
per
at
ive
sc
hem
e
BF(CS
-
S
IC
-
BC
H)
is
bette
r
than
Un
c
oded
-
Nonc
oope
rati
ve
syst
e
m
[19]
.
Mo
re
sp
eci
fical
ly
,
to
achieve
a
BE
R
of
a
bout
10
−
4
the
require
d
S
N
R
for
the
pro
pos
ed
sc
hem
e
BF(CS
-
S
IC
-
BC
H
)
is
a
bout
10
dB
le
ss
th
an
t
he
Un
c
oded
-
N
on
c
oope
rati
ve
syst
e
m
[19]
.
Also,
the
per
f
orm
ance
of
beam
fo
rm
ing
of
base
d
on
S
LR
in
SI
C
cooper
at
ive
sc
hem
e
BF(CS
-
S
IC
-
BC
H
)
is bett
er tha
n U
ncode
d
-
c
oope
r
at
ive s
chem
e BF(CS
-
SI
C
)
as
s
how
n
in
in
Tab
le
3
.
Figure
3. BER
perform
ances o
f
the c
oded
and
Un
c
oded
pr
opose
d s
chem
e and No
ncoo
perat
ive syst
em
for
M =
4 N=
5,
i
n
Ri
ci
an
i
nter user
ch
a
nnel
SN
R
=
20
with
k
=
25 a
nd β
=
0
.
1
Table
3.
Desc
ribe of
res
ult i
n
Figur
e
3
Para
m
eters
BER co
m
p
ass
io
n
Poin
t
10
−
4
BF(CS
-
S
IC)
T=
4
R=5
R
eq
u
ired SNR
3
BF(CS
-
S
IC
-
BC
H)
T=4 R=5
R
eq
u
ired SNR
-
7
BF
Co
d
ed
-
No
n
co
o
p
erative T
=4
R=5
R
eq
u
ired SNR
-
7
.
5
BF
Un
co
d
ed
-
No
n
co
o
p
erative T
=4
R
=6
[
1
9
]
R
eq
u
ired SNR
5
BF
Un
co
d
ed
-
No
n
co
o
p
erative T
=4
R
=5
[
1
9
]
R
eq
u
ired SNR
6
.5
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
Maximizi
ng
sig
na
l t
o
le
ak
age
ra
ti
os
i
n
M
IM
O
BC
H
c
ooper
ative
b
e
am
f
ormin
g
sc
he
me
(
Mo
hamm
e
d Fa
dh
il
)
3709
Figure
4,
s
ho
ws
the
c
om
pari
so
n
betwee
n
Cod
e
d
-
N
on
c
oo
per
at
ive
syst
em
and
the
pro
po
s
ed
sc
hem
e
BF(CS
-
SI
C
-
B
CH)
to
sel
ect
the
opti
m
u
m
value
of
β.
w
here
a
base
sta
ti
on
(BS
)
e
qu
i
pped
wit
h
M=
4
a
nten
na
s
com
m
un
ic
at
es
with
N
=4
in
e
ach
us
e
r.
W
e
c
hoos
e
th
ree
dif
fer
e
nt
value
β
for
the
pro
pos
ed
schem
e,
to
m
ade
com
par
ison
w
it
h
Cod
e
d
-
Noncoo
per
at
ive
s
yst
e
m
,
first
wh
en
β
=
0.9
,
the
per
f
orm
a
nce
of
the
Cod
e
d
-
Nonc
oope
rati
ve
syst
e
m
,
wh
ic
h
m
axi
m
iz
es
t
he
us
e
fu
l
po
we
r
of
use
rs
a
nd
neg
le
ct
s
the
m
ulti
-
us
e
r
interf
eren
ce
,
has
bette
r
perf
or
m
ance
tha
n
t
he
propose
d
sc
hem
e
BF(CS
-
S
IC
-
BC
H
).
T
hat
is
beca
us
e
t
he
eff
ect
of
m
ulti
-
us
e
r
interfe
ren
ce
is
high,
w
hich
be
com
es
the
m
a
i
n
f
act
or
lim
it
i
ng
syst
em
per
f
or
m
ance.
T
herefo
re,
i
n
the
s
econd
ti
m
e
slot,
the
us
ers
will
sh
a
re
the
sign
al
s
with
hi
gh
i
nterf
e
r
ence
val
ue.
W
hile
in
the
sec
ond
sce
nar
i
o,
wh
e
n
β
=
0.7
an
d
0.5
the
e
ff
ect
of
th
e
inter
fer
e
nce
on
s
har
i
ng
sig
nals
is
reduce
d.
T
her
e
fore,
th
e
pe
rfor
m
ance
of
th
e
pro
po
se
d
sche
m
e
will
be
im
pro
ved
but
sti
ll
worse
tha
n
t
he
C
od
e
d
-
N
on
coope
rati
ve
sy
stem
.
W
hilst
f
or
the
third
sce
na
rio,
wh
e
n
β
=
0.1,
no
ise
is
th
e
m
ai
n
facto
r
li
m
iting
the
syst
em
perform
ance.
That
is
beca
use
the
pro
po
se
d
sc
he
m
e
BF(CS
-
SIC
-
BC
H)
is
m
akin
g
the
inte
rf
e
ren
ce
si
gn
a
ls
turn
int
o
use
fu
l
sig
nals
wh
e
n
it
detect
ed
these
sign
al
s
by
us
i
ng
SI
C,
an
d
it
cou
l
d
get
the
be
nef
it
from
the
m
ul
ti
-
us
er
inte
rf
e
ren
ce
via
s
ha
rin
g
these
sig
nals
am
on
g
us
e
rs.
T
her
e
fore,
the
pe
rfor
m
ance
of
the
pro
po
se
d
schem
e
BF(CS
-
S
IC
-
BC
H
)
will
be
i
m
pr
oved
as s
how
n
i
n
in
Ta
ble
4
.
Figure
5,
pr
es
ents
the
perfor
m
ance
of
pro
pose
d
schem
e
BF(CS
-
SI
C
-
B
CH)
in
dow
nlink
c
ha
nn
el
s
betwee
n
the
BS
and
the
us
e
r
s
wh
ic
h
hav
e
equ
al
val
ue
of
the
SN
R
in
t
he
cha
nn
el
,
w
hile
the
SN
R
of
the
inter
-
us
er
c
hannel
equ
al
to
5,
and
25
dB
.
Th
e
resu
lt
sh
ows
t
he
perform
ance
of
the
syst
em
is
enh
ance
d
wh
e
n
the inter
-
c
ha
nnel
S
NR inc
rea
se.
Figure
4. BER
perform
ances o
f
the
pro
pose
d
sc
hem
e
and Co
de
d
-
Nonc
oo
per
at
ive
s
yst
e
m
f
or
M =
4
a
nd
N
=
4
i
n
Ri
ci
an
inte
r user
ch
a
nn
el
with
k=
25 and
SN
R
=
20
Figure
5. BER
perform
ances o
f
the
pro
pose
d
a
nd
ancho
r
sc
hem
e
for
M =
4
N=
4,
i
n
Ri
ci
an
i
nter user
channel
with
k = 2
5
a
nd β =
0.
1
Table
4.
Desc
ribe of
res
ult i
n
Figure
4
Para
m
eters
BER co
m
p
ass
io
n
Poin
t
10
−
4
BF(CS
-
S
IC
-
BC
H
)
T=4
R=4
B=0
.1
R
eq
u
ired SNR
1
BF(CS
-
S
IC
-
BC
H)
T=4 R=4
B
=0
.5
R
eq
u
ired SNR
2
BF(CS
-
S
IC
-
BC
H)
T=4 R=4
B
=0
.7
R
eq
u
ired SNR
2
.3
BF(CS
-
S
IC
-
BC
H)
T=4 R=4
B
=0
.9
R
eq
u
ired SNR
3
BF
Co
d
ed
-
No
n
co
o
p
erative T
=4
R=4
R
eq
u
ired SNR
2
Figure
6,
s
hows
the
syst
em
perform
ance
wh
e
n
the
inter
-
us
e
r
c
ha
nnel
us
e
d
a
li
ne
of
sigh
t
L
O
S
env
i
ronm
ent
(o
ve
r
a
correla
te
d
reali
sti
c
R
ic
ia
n
fad
in
g
channel)
.
W
he
r
e
the
per
f
or
m
ance
of
the
propose
d
schem
e
BF(CS
-
S
IC
-
BC
H
)
is
bette
r
t
ha
n
the
pe
rfo
r
m
ance
of
th
e
Co
ded
-
N
on
coope
rati
ve
s
yst
e
m
.
More
s
pecifica
ll
y,
in
case
k
=
2
5
to
ac
hie
ve
a
BER
of
about
10
−
5
the
require
d
SN
R
f
or
the
pro
pose
d
schem
e
BF(CS
-
SI
C
-
BC
H)
is
about
1
dB
le
s
s
t
han
Co
ded
-
Nonc
oope
rati
ve
syst
e
m
.
It
a
lso
show
s
the
pr
opos
e
d
schem
e
BF
(CS
-
SI
C
-
BC
H
)
perform
ance
is
worse
tha
n
the
Co
de
d
-
Nonc
oope
rati
ve
syst
e
m
w
hen
k
is
decr
ease
d.
In
oth
er
w
ords
,
wh
e
n
the
in
te
r
-
us
e
r
cha
nnel
LOS
is
r
edu
ce
d
the
to
ta
l
pr
opos
e
d
syst
e
m
perform
ance
will
a
lso
reduce.
It
is
necessary
fo
r
these
us
e
rs
to
identify
a
su
it
able
par
tn
er
to
ob
ta
in
optim
a
l
perform
ance
thr
ough
kn
ow
le
dg
e
of
the
in
te
r
-
us
er
c
ha
nn
el
char
act
e
risti
cs
betwee
n
each
us
er
a
nd
it
s’
pa
rtne
r
as sho
wn in
i
n Table
5
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
20
19
:
3
7
0
1
-
3
7
1
3
3710
Figure
6. BER
perform
ances o
f
the
pro
pose
d
a
nd Co
de
d
-
N
on
c
oope
rati
ve sy
stem
f
or
M
= 4
N=
4, in
L
OS
env
i
ronm
ent o
f
inter
us
e
r
c
hannel S
NR=2
0, β
= 0.1
Table
5.
De
sc
ribe of
res
ult i
n
Figure
6
Para
m
eters
BER co
m
p
ass
io
n
Poin
t
10
−
4
BF(CS
-
S
IC
-
BC
H
)
T=4
R=4
KdB
=2
5
R
eq
u
ired SNR
1
BF(CS
-
S
IC
-
BC
H)
T=4 R=4
Kd
B=1
5
R
eq
u
ired SNR
4
BF(CS
-
S
IC
-
BC
H)
T=4 R=4
Kd
B=1
0
R
eq
u
ired SNR
2
a
t
10
−
3
BF
Co
d
ed
-
No
n
co
o
p
erative
T=4
R=4
R
eq
u
ired SNR
2
Figure
7,
sho
ws
com
par
iso
n
betwe
en
B
ER
per
f
orm
ance
BF(CS
-
S
I
C
-
BC
H)
a
nd
the
Cod
e
d
-
Nonc
oope
rati
ve
syst
e
m
e
m
plo
ye
d
m
ulti
-
antenn
a
in
both
Bs
an
d
us
er
s,
wh
e
re
BS
a
ntenn
a
s
M
=
2,4
and
6
wh
il
e
us
er
s
a
nt
enn
as
U
=
2,4
an
d
6,
the
res
ult
s
ho
ws
t
he
syst
e
m
has
a
s
ign
ific
a
nt
im
pr
ov
em
ent
f
or
M
=
6
,
N=6
i
n
both
Cod
e
d
-
N
on
c
oo
per
at
ive
syst
e
m
and
pro
pos
ed
schem
e
BF(CS
-
S
IC
-
BC
H
),
w
hile
the
pr
opos
e
d
schem
e
BF(CS
-
SI
C
-
BC
H)
sti
l
l
go
t
en
han
c
e
m
ent
com
par
ing
with
C
od
e
d
-
Nonc
oope
rati
ve
syst
e
m
as
sh
own
in
in Ta
ble
6
.
Figure
7
.
BER
perform
ances o
f
the
pro
pose
d
a
nd Co
de
d
-
N
on
c
oope
rati
ve
syst
e
m
f
or
M
=2,
4
a
nd
6,
N
=
2, 4 an
d 6, i
n
Ri
ci
an
i
nter
u
se
r
c
hannel
with
k
=
25
with S
NR=2
0, β =
0
.
1
Table
6.
Desc
ribe of
res
ult i
n
Figure
7
Para
m
eters
BER co
m
p
ass
io
n
Poin
t
10
−
4
BF(CS
-
S
IC
-
BC
H
)
T=2
R=2
R
eq
u
ired SNR
3
BF
Co
d
ed
-
No
n
co
o
p
erative T
=2
R=2
R
eq
u
ired SNR
3
.5
BF(CS
-
S
IC
-
BC
H)
T=4 R=4
R
eq
u
ired SNR
1
BF
Co
d
ed
-
No
n
co
o
p
erative T
=4
R=4
R
eq
u
ired SNR
2
BF(CS
-
S
IC
-
BC
H)
T=6 R=6
R
eq
u
ired SNR
2
.5
BF
Co
d
ed
-
No
n
co
o
p
erative T
=6
R=6
R
eq
u
ired SNR
2
Evaluation Warning : The document was created with Spire.PDF for Python.