TELKOM
NIKA
, Vol.14, No
.4, Dece
mbe
r
2016, pp. 13
90~139
6
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v14i4.4052
1390
Re
cei
v
ed Ma
y 26, 201
6; Revi
sed Septe
m
ber
23, 201
6; Acce
pted
Octob
e
r 8, 20
16
Multi-Stage Partial Parallel Interference Cancellation
Algorithm for MUSA Systems
Yan Liang, Han Wu*, Gu
a
n
g
y
u Wang
Cho
ngq
in
g Ke
y L
abor
ator
y
of
Mobil
e
Comm
unic
a
tion T
e
ch
nol
og
y
Cho
ngq
in
g Uni
v
ersit
y
of Posts
and T
e
lecom
m
unic
a
tions (C
QUPT
), Chong
qin
g
, 400
065,
Chin
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
:
w
u
ha
n0
502
3
@
16
3.com
A
b
st
r
a
ct
Multi-User
Sh
a
r
ed Acc
e
ss is
a n
on-orth
og
o
nal
mu
lti
p
le
ac
cess sch
e
m
e
of 5G,w
hich h
a
s a
hi
gh
computati
o
n
a
l
compl
e
xity a
n
d
a l
a
rge
ti
me
d
e
lay
du
e to
th
e
usa
ge
of succ
essive
interfer
ence
canc
ell
a
ti
on
detectio
n
al
gor
ithm. T
h
is
pap
er prop
oses a mu
lti-stag
e
par
tial
par
al
lel
i
n
t
e
rferenc
e canc
ellati
on
detecti
o
n
alg
o
rith
m, w
h
ic
h do
es n
o
t req
u
ire r
epe
ated
orderi
ng
an
d r
epe
ated
matrix
inversi
on. In t
he first stage
of
d
e
t
e
c
ti
on
, th
e
b
i
ts o
f
th
e
s
e use
r
s wi
th
go
od ch
an
ne
l
co
n
d
i
t
ions
w
i
l
l
b
e
o
u
tputted, an
d
t
he influ
enc
e o
f
mu
ltipl
e
acces
s
interfer
ence
on users
w
i
th bad
cha
n
n
e
l c
ond
itions
i
n
th
e sec
ond
stag
e w
ill
be
decr
e
ased
.
T
heoretic
al a
n
a
lysis a
nd si
mulati
on res
u
lts show
that
the symb
o
l err
o
r rate of the pro
p
osed a
l
g
o
rith
m is
slightly
better t
han t
hat of th
e
tw
o-stage MM
SE-PIC, and th
e co
mp
lexity
is
reduc
ed. In th
e
mea
n
w
h
il
e, the
computati
o
n
a
l
compl
e
xity is s
i
gnific
antly r
e
d
u
c
ed w
i
tho
u
t SER perfor
m
a
n
c
e de
grad
atio
n
w
hen co
mp
ar
ed
with MMSE-SIC algorithm
.
Ke
y
w
ords
:
Multi-User
Sh
ared
Access,
no
n-orth
ogo
nal
m
u
ltiple access,
succ
essive interferenc
e
cance
llati
on, p
a
rall
el i
n
terfere
n
ce
canc
ell
a
tio
n
, sym
bol error
rate
Copy
right
©
2016 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
In the mobile communi
cati
on system [1]
,
multip
le access techni
qu
e is to allow use
r
s to
sha
r
e the sca
r
ce
radi
o re
source
s and
communi
cate
simultan
eou
sl
y. From the first to the fourth
gene
ration m
obile commu
nicatio
n
, FDMA, TDMA, CDMA a
nd
OFDMA a
r
e
use
d
as th
e main
multiple a
c
ce
ss sch
e
me
s,
all of the
m
a
r
e o
r
thogo
nal multiple
a
c
ce
ss. Ho
weve
r, in
o
r
de
r
to m
eet
the deman
ds of massive con
n
e
c
tion
s, high sp
ect
r
u
m
efficiency,
and high ca
pacity and lo
w
latency [2-3] of the fifth generatio
n mobi
le comm
uni
cation [4-5], n
on-o
r
tho
gon
a
l
multiple access
has attracte
d
a lot of consid
eratio
ns.
So far,
there are th
ree
well-kn
own p
r
opo
sal
s
in t
h
e
comm
unity of non-ortho
g
o
nal multiple
acce
ss
sc
he
mes in
Chin
a: ZTE Corp
oration
pro
p
o
s
e
s
Multi-User Shared Access (MUS
A
)
[6], which a
c
hie
v
es free sch
edulin
g tran
smissi
on; Spa
r
se
Cod
e
Multipl
e
Acce
ss (S
CMA) [7] from Hua
w
ei
Co
rpo
r
ation, ha
s reali
z
e
d
the chann
el overload
by 300%; Pattern Divisi
on
Multiple Acce
ss (P
DM
A) [8
], supporte
d by Datang Te
lecom, redu
ces
the reali
z
atio
n compl
e
xity.
The
key el
e
m
ents
of M
U
SA system i
n
clud
e
compl
e
x multi-dom
ai
n spre
adin
g
cod
e
s [9
-
10] an
d a
d
va
nce
d
su
ccessive interfe
r
en
ce
ca
ncellatio
n
(SIC)
re
ceiv
er [1
1]. The
sprea
d
ing
cod
e
s
among
vari
o
u
s
users a
r
e
non
-o
rthog
o
nal in
MUSA
syste
m
, therefore th
e p
e
rforman
c
e
of
the
MUSA systems will be m
a
inly affected by multiple
access interfer
ence
(MAI), also includi
ng
multipath inte
rfere
n
ce a
nd
noise. In o
r
d
e
r
to
meet
th
e
requi
rem
ents
of the fifth g
e
neratio
n m
obi
le
comm
uni
cati
on
su
ch
a
s
massive
co
n
nectio
n
s,
hig
h
spe
c
trum
e
fficiency, hi
g
h
capa
city a
n
d
lo
w
latency, ho
w
to increa
se t
he capa
city and e
limi
nate
the MAI are
major
chall
e
nge
s for M
U
SA
sy
st
em
s.
One of th
e
key te
chnol
o
g
ies to
re
du
ce the
MAI
of MUSA
system i
s
the
multi-u
s
er
detectio
n
alg
o
rithm i.e. minimum me
an sq
ua
re e
rro
r-su
cce
ssi
ve interfere
n
c
e can
c
ellati
on
(MMSE-SIC) algorith
m
.
Users are so
rte
d
in
as
ce
ndi
ng order
accordin
g to the
SINR of use
r
s,
then the stro
nge
st use
r
wi
ll be firstly detected,
and the influen
ce
corre
s
p
ondin
g
to the detected
use
r
will be subtra
cted
fro
m
the re
ceive
d
si
gnal.
Si
milar step
s su
ch
a
s
so
rting a
nd
d
e
tectin
g will
be
repe
ated
until all
users are d
e
tecte
d
.
It is o
b
viou
s that the
later dete
c
tion
ord
e
r
of the
user is,
the higher degree accuracy of
the detection
will be. However,
the algorithm
is an iterati
v
e
pro
c
ed
ures i
n
clu
d
ing
ord
e
ring,
estim
a
tion an
d
su
btractio
n,
wh
ich req
u
ire compl
e
x
mat
r
ix
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Multi-stage partial parallel
interferen
ce cancellation al
gorithm
for MUSA… (Yan
Liang
)
1391
comp
utation
e.g. matrix in
versio
n d
ue t
o
the
MMSE
crite
r
ion.
Un
fortunately, the u
s
e
r
nu
m
ber
deci
de the nu
mber of iterati
v
e
procedu
re
s, therefo
r
e the MMSE-SI
C dete
c
tion al
gorithm result in
a high
com
p
l
e
xity and a la
rge p
r
o
c
e
s
sin
g
delay of th
e syste
m
wh
en the u
s
e
r
n
u
mbe
r
is
getting
large.
A
s
is
well-
kno
w
n,
t
he cla
ssi
cal
MM
SE-PIC a
l
gorithm, whi
c
h ha
s bee
n
widely use
d
in
CDMA and MIMO
syste
m
s,
ha
s seve
ral
cha
r
a
c
teri
st
ics such a
s
low
comp
uta
t
ional comple
xity
and
a small
delay. In ge
neral, th
e M
M
SE-PIC alg
o
rithm
req
u
ires m
u
lti-stag
e structu
r
e,
an
d
detect
s
all
users’
si
gnal
s i
n
pa
rallel.
Ho
wever at
the ca
se
of parall
e
l
proc
essing
, those
stro
n
ger
use
r
s
always
influent we
aker u
s
e
r
s, whi
c
h raise t
he i
m
pact of MAI
.
In terms of the short
c
omi
ngs
of the MMSE-SIC and t
he traditio
nal
MMSE-PIC
algorithm,
a multi-sta
g
e
partial pa
ra
llel
interferen
ce
can
c
ell
a
tion algorith
m
is
prop
osed, which o
n
ly needs o
n
ce orderin
g and t
he
numbe
r of matrix inversi
o
n with re
spe
c
t to t
he MMSE-SIC has
si
gnifica
ntly reduced. Theo
retical
analysi
s
and
simulatio
n
re
sults
sho
w
th
at
the co
m
p
le
xity of the propo
sed
dete
c
tion alg
o
rithm
is
signifi
cantly redu
ced
with
o
u
t SER
perfo
rman
ce
de
gradation
when
co
mpa
r
ed
with MMSE-SI
C
algorith
m
.
2. Sy
stem Model
Figure 1 describ
es the up
link MUSA system archite
c
ture.
A
s
sum
i
ng that there are K
use
r
s,
ea
ch
use
r
’s d
a
ta
are
spre
ad
by dedi
cate
d
co
mplex
m
u
lti-dom
ain
sprea
d
ing
cod
e
s
respe
c
tively.
Then all spre
ad symbol
s a
r
e tran
smitte
d
over the sam
e
time-freq
u
e
n
cy re
sou
r
ce
s.
1
()
H
2
()
H
()
K
H
Figure 1. Upli
nk MUSA System
The re
ceive
d
sign
al after the cha
nnel
ca
n be rep
r
e
s
e
n
ted by:
1
K
kk
k
k
rg
s
x
z
=
.
(1)
Whe
r
e
k
x
is the
transmitted symbol of user k,
k
s
is the
spre
ading
seq
uen
ce of u
s
e
r
k,
k
g
is the
cha
nnel
gain
of u
s
er k,
a
nd
z is a
co
mplex-valu
ed
noi
se ta
ken
from
a zero
mean
Ga
ussian
distrib
u
tion wi
th variance
2
.
The re
ceive
d
sign
al ca
n be
rewritten by vector a
s
:
rH
x
z
(2)
Whe
r
e
12
(,
,
,
)
T
N
rr
r
r
,
12
(,
,
,
)
T
K
x
xx
x
,
12
(,
,
,
)
T
N
zz
z
z
,
2
~(
0
,
)
zI
CN
.
H
is
the c
h
annel matrix, and
k
h
in the
kth
co
lumn of
H
is e
qual to
kk
gs
. At the receiving
end,
MMSE-SIC receive
r
i
s
u
s
ed
to dem
odulate
a
n
d
re
cove
r th
e
data
of e
a
c
h
user fro
m
the
s
u
pe
r
i
mp
os
ed
s
y
mb
o
l
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 4, Dece
mb
er 201
6 : 1390 – 139
6
1392
3. Multi-user Detection
Algorithm
3.1. MMSE-SIC Detec
t
ion Algorithm
The MMSE
-
SIC algo
rith
m is
ba
sed
on the
sc
he
me of
su
cce
ssive
processing,
whi
c
h
eliminate
s
int
e
r-i
nterfe
ren
c
e ste
p
by ste
p
, and
e
a
ch
step only det
ects
on
e use
r
.
The
r
efo
r
e, if
there
are K
use
r
s in
the
system, th
ere shoul
d
b
e
K
time’s det
ection
s.
Fi
gu
re 2 sho
w
s
t
he
stru
cture of the MMSE-SIC.
~~
1
1
1
y
yh
x
~~
~
2
2
21
yy
h
x
~~
~
3
3
32
yy
h
x
~
3
y
~
2
y
~
1
y
~
2
x
~
1
x
Figure 2. MMSE-SIC Structure
The main
ste
p
s of the MM
SE-SIC algori
t
hm are a
s
follows
:
Step 1: Initialization
11
1,
,
ir
r
H
H
H2
1
H
(
GH
H
σ
I
H
)
Step 2: SIC
For i=1: K
Orde
rin
g
:
11
{,
,
}
ar
g
m
ax
(
)
i
ij
jk
k
kS
I
N
R
Nulling
ve
ctor
:
()
ii
kk
ω
G
Nulling
:
ii
kk
i
y
ω
r
Har
d
de
cisi
on
:
~
()
ii
kk
x
Qy
SIC
:
~
1
()
ii
ii
k
k
rr
H
x
Upd
a
te the chann
el matrix
:
~
1
()
ii
ii
k
k
rr
H
x
Cal
c
ulate the
weig
ht matrix:
H2
1
H
11
1
1
(
σ
)
ii
i
i
GH
H
I
H
End
Whe
r
e Q
(
.) a
nd I re
spe
c
tively denote th
e quanti
z
atio
n (sli
cin
g
) o
p
e
ration
and id
entity matrix,
i
k
H
denote
s
th
e
matrix ge
nerated by
delet
ing the
k-th
column
of H.
()
k
H
denote
s
th
e k-colu
mn of
H
.
H
denote
s
the
Frobe
niu
s
no
rm of matrix
H
.
The SINR i
s
given by:
2
i
22
2
il
i
i
i
li
Gh
SI
N
R
Gh
G
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Multi-stage partial parallel
interferen
ce cancellation al
gorithm
for MUSA… (Yan
Liang
)
1393
There
will be
K time’s
ma
trix inversion
and K
-
1 tim
e
s
ord
e
rin
g
a
nd
when
the
r
e are K
use
r
s in th
e
system, therefore, the
com
p
lexity is
hig
h
and
the d
e
la
y is la
rge. A
s
for the
dem
a
nds
of massive connection of
5G, those will
be worse.
3.2. Propose
d
Algorithm
The propo
se
d MPPIC algorithm mai
n
ly
adapts two stage
s M
M
SE-PIC co
nstru
c
tion
instea
d of MMSE-SIC because of con
s
i
deratio
n
of complexity and pro
c
e
ssi
ng
delay. Comp
ared
with tradition
al MMSE-PIC, propo
sed al
gorithm di
vid
e
first stage d
e
tected u
s
e
r
s into two grou
ps,
stronger
users will be out
puts and
weaker users will
be
passed to next stage detection.
In
gene
ral, users are
sorte
d
in ascend o
r
der a
c
cordi
n
g to their ch
annel stat
es,
then MMSE-PIC
algorith
m
is adopte
d
in the first-stage
detectio
n
.
Interferen
ce of all use
r
s
will be re
con
s
tru
c
ted
according to
the output
s o
f
the first-sta
ge det
e
c
tion
and the
ch
an
nel e
s
timatio
n
, and the
n
the
bits of these
stronger users
w
ill be outputted. Final, t
he remai
n
ing users
are det
ected
again with
the MMSE-PIC algo
rithm. Figure 3 pre
s
ents the sc
h
e
m
atic diag
ra
m of the prop
ose
d
algo
rith
m.
Figure 3. The
Schemati
c
Diagra
m
of the Propo
se
d Algorithm
The m
a
in
ste
p
s
of the
two
-
stage
pa
rtial
parall
e
l inte
rf
eren
ce
can
c
e
llation al
gorit
hm a
r
e
as
follows
:
Step 1: Initialization
H2
1
H
(
σ
)
(
M
M
S
E
)
GH
H
I
H
~~
~
1
[
(
0),
.
..,
(
0
)]
K
xG
r
x
x
Orde
rin
g
:
arg
m
ax
ii
kH
Step 2: the first-stage PIC
For k=1: K
^~
~
~
~
1
11
[
(
0
)
,
,
(0
)
,
0
,
(0
)
,
,
(
0
)
]
kK
kk
xx
x
x
x
PIC:
^
k
k
rr
H
x
Har
d
De
ci
sio
n
:
~
(1)
(
)
k
kk
x
QH
r
End
Step 3: the seco
nd-stag
e PIC
Acco
rdi
ng to the ord
e
ri
ng, use
rs
w
ith larger
i
H
will be outputted.
The remaining users
will be detec
ted again with PIC algorithm.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 4, Dece
mb
er 201
6 : 1390 – 139
6
1394
The numb
e
r of matrix inversi
on an
d
user
o
r
de
ri
ng is sig
n
ificantly red
u
ced whe
n
comp
ared wit
h
MMSE-SIC algorithm, especi
a
lly when
the number
of users is large. In addition,
as the
PIC
detecto
r [12]
adopt
s the
interferen
ce
can
c
ell
a
tion i
n
pa
rallel, a
nd the influ
e
n
ce
cau
s
e
d
by the others use
r
s is
al
way
s
existed, whi
c
h affects
the
perform
an
ce
of the system.
While th
e propo
sed
algo
rithm output
s the strong
er
use
r
s in th
e first
stage,
whi
c
h in
ret
u
rn
redu
ce the M
A
I of the weaker u
s
e
r
s in the se
co
nd st
age.
The p
r
o
p
o
s
e
d
algo
rithm i
s
simila
r to
the traditio
n
a
l
dete
c
tion a
l
gorithm, t
w
o
-
stag
e
MMSE-PIC []. However, the main differenc
e is
t
hat the propos
ed
algorithm
outputs
the
s
t
ronger
use
r
s in the
first dete
c
tion
stage,
whi
c
h re
du
ce the
MAI of the
wea
k
e
r
u
s
e
r
s in the
se
co
nd
detectio
n
. While the traditi
onal two
-
stag
e MMSE-PI
C does
not out
put sign
als i
n
the first stag
e,
and the
MAI
whi
c
h i
s
cau
s
ed
by the
s
e
stro
nge
r u
s
e
r
s
ha
s a l
a
rg
e influen
ce
o
n
the
s
e
wea
k
er
users. In addi
tion, the in
fluence is al
way
s
exi
s
ted, thus the
weaker
users will
suff
er large MAI i
n
the second
d
e
tection.
The
r
efore, the
r
e
will b
e
p
e
rfo
r
mance l
o
ss
whe
n
u
s
in
g t
he traditional
two-
stage
MMSE-PIC alg
o
rithm
.
What’
s
m
o
re, as the
stro
nger u
s
e
r
s a
r
e outp
u
tted in
the first
stag
e,
whi
c
h in retu
rn redu
ce the
compl
e
xity of
the prop
osed
algorith
m
.
4. Performan
ce and Complexit
y
Analy
s
is
In this se
cti
on, the
pe
rforma
nce a
n
d
co
mplexity of the
pro
posed
algo
ri
thm are
discu
s
sed. F
o
r convenie
n
ce, the p
r
opo
sed al
go
rithm is lab
e
led a
s
MPPIC. Simulation
para
m
eters a
r
e a
s
sho
w
n
in Table
1. The lengt
h of spreadi
ng
co
des i
s
N, an
d the num
ber of
use
r
s i
s
K.
4.1. Perform
a
nce Analy
s
is
In this subs
ec
tion, the
s
y
mbol
error rate
(SE
R
) pe
rforma
nce of
the p
r
o
posed
algorith
m
,
two-s
t
age MMSE-PIC and MMSE-SIC algorithm is
comp
ared. Fi
gure
4
and
F
i
gure
5
sh
ow the
SER perfo
rm
ance re
sult
s of the three
algorith
m
s
wi
th different n
u
mbe
r
of use
r
s a
nd differe
nt
length of sp
re
ading
cod
e
s.
The othe
rs p
a
ram
e
ters are sho
w
n in T
able 1.
Figure 4. SER Perfo
r
man
c
e of the thre
e
Algorithm for
N=8, K=4
Figure 5. SER Perfo
r
man
c
e of the Three
Algorithm for
N=16, K=10
Table1. Simul
a
tion Param
e
ters
Parameters
Value
N
K
Channel model
Channel Estimation
SNR
Modulation
8,16
4,10
Ra
y
l
eigh fading c
hannel
Ideal
0-14dB
Q
PSK
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Multi-stage partial parallel
interferen
ce cancellation al
gorithm
for MUSA… (Yan
Liang
)
1395
As are sho
w
n
in Figure 4 a
nd Figure 5, the
SER perfo
rman
ce of the
MPPIC algorithm is
comp
arable t
o
the MMSE-SIC, but it is better t
han
that of the two-sta
ge MMS
E-PIC, with t
he
cha
nge of K
and
N, whi
c
h
is du
e to the receive
d
sig
n
a
l
throug
h the
MMSE detect
o
r first, and t
he
MMSE detect
o
r ma
ke
s a
compromise b
e
twee
n noi
se
and MAI. Wh
at’s mo
re, the
stro
nge
r u
s
e
r
s
have bee
n o
u
tputted in th
e first dete
c
ti
on stag
e,
whi
c
h redu
ce th
e MAI of the wea
k
e
r
u
s
ers in
the second
stage. T
h
u
s
, the wea
k
e
r
users h
a
ve
more a
c
curate dete
c
tion
, whi
c
h in
re
turn
improve
s
the
system p
e
rfo
r
mance.
4.2. Comple
xit
y
Analy
s
is
The complexi
ty of
the thre
e algorith
m
s
is anal
y
z
ed i
n
this su
bsection, we co
nsider all
compl
e
x mul
t
iplication
s
/di
v
ision
s
an
d
compl
e
x
ad
d
i
tions/subtra
ctions and use
floating po
int
operation (flo
p) as u
n
it to measure com
p
lexity
. One multiplicatio
n and on
e addi
tion respe
c
tively
corre
s
p
ond to
6 flops and 2
flops [13].
For a
mn
ch
an
nel matrix, the calculatio
n of t
he wei
ght matrix of
MMSE requ
ires
22
52
nm
n
m
n
multiplication
s
and
22
52
3
2
2
3
2
nm
n
m
n
n
additions [14].
Table 2. Co
m
pari
s
on of the
Complexity
Algorithm Multiplicat
ions
Additions
Flops
K=4
N=8
MMSE-SIC
1106
804
8244
T
w
o-stage MMS
E-PIC
800
596
5992
MPPIC 620
453
4626
K=10
N=16
MMSE-SIC
24393
21090
188538
T
w
o-stage MMS
E-PIC
8760
7710
67980
MPPIC 6075
5264
46978
Table 2
sho
w
s the compl
e
xity of the three
algo
rithm
s
with differe
nt N and K. Th
e MMSE-
SIC has the
highe
st co
mp
lexity and the pro
pos
ed
algorith
m
ha
s the lowest
compl
e
xity. T
he
compl
e
xity of the MMSE-SIC in
crea
ses
sha
r
pl
y
with the in
crease of N a
nd K, whil
e
the
compl
e
xity of
the MPPIC is increa
sed
rel
a
tively slow.
As
the MMS
E-SIC algorit
h
m involves
K-1 ti
me
s o
r
derin
g a
nd K
times matrix
inversion
whe
n
there
a
r
e K users acce
ss to the
system, wh
ile
the pro
p
o
s
ed
algorithm o
n
l
y
requires o
n
c
e
orde
rin
g
and
two times matrix inversi
on , and the
matrix inversion h
a
s a l
a
rge
com
p
le
xity.
What’
s
mo
re
, stron
ger u
s
ers
are
out
putted in
the
first dete
c
ti
on sta
ge, which
red
u
ce
the
compl
e
xity when
com
pare
d
with th
at of
the
two
-
stag
e MMSE-PIC. Therefore,
MMSE-SIC h
a
s
the large
s
t co
mplexity, whil
e the p
r
o
p
o
s
ed alg
o
ri
thm
has the lo
we
st complexity, espe
cially when
the numbe
r o
f
acce
ss users is large in
the massive conne
ction
scenari
o
of 5G.
5. Conclusio
n
In this p
ape
r,
a multi-stag
e pa
rtial pa
ra
lle
l interfe
r
en
ce
can
c
ell
a
tion multiu
se
r
detectio
n
algorith
m
, M
PPIC, is
pro
p
o
se
d fo
r M
U
SA system
s.
It requi
re
s l
e
ss mat
r
ix inv
e
rsi
o
n
an
d u
s
ers
orde
rin
g
wh
e
n
com
pared
with MMSE-SIC al
go
rith
m, which ca
n achi
eve n
ear MMSE-S
I
C
perfo
rman
ce
while the com
p
lexity is sign
ificantly
redu
ced, esp
e
ci
ally in the massi
ve conn
ectio
n
s
scena
rio
of the fifth gene
rat
i
on mo
bile
co
mmuni
cati
on,
whe
n
the
nu
mber
of u
s
ers and th
e len
g
th
of the sp
re
ad
ing code
s a
r
e larg
e. In a
ddition,
the S
E
R pe
rform
a
nce
and it
s
complexity of the
MPPIC algori
t
hm are b
e
tter than that of
the
two-stag
e MMSE PIC algorithm. A
s
the p
r
opo
sed
algorith
m
ad
o
p
ts the PIC structu
r
e, the
compl
e
xi
ty is low.
What’
s
more, the
stronge
r u
s
e
r
s
are
outputted in the first stag
e
detection, which be
nef
its
the wea
k
e
r
u
s
ers in the seco
nd stag
e as
the wea
k
e
r
u
s
ers suffer l
e
ss MAI. The
r
efore,
it
ca
n
guarantee
th
e dete
c
tion
p
e
rform
a
n
c
e
a
nd
kee
p
a low
co
mplex
i
t
y
.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 4, Dece
mb
er 201
6 : 1390 – 139
6
1396
Ackn
o
w
l
e
dg
ements
This work i
s
su
ppo
rted
by Nationa
l Science a
nd Technol
o
g
y Major Project of
Chin
a
(20
16Z
X03001
010
)
and th
e Sci
e
nce
an
d Te
chnolo
g
y Re
search P
r
oje
c
t
of Chon
gqin
g
Munici
pal Ed
ucatio
n Com
m
issi
on of Ch
ina (KJ140
04
37).
R
e
fe
re
nc
es
[1]
David T
s
e, Pramod V
i
s
w
a
n
a
t
h. F
undame
n
t
a
ls of
W
i
rel
e
s
s
Commun
i
cat
i
on. C
a
mbri
dg
e Univ
ersit
y
Press. 2005.
[2]
Z
hao Guof
en
g, Che
n
Ji
ng
, Han Yu
an
bi
ng. Pr
osp
e
cti
v
e net
w
o
rk techn
i
qu
es for
5G mobi
l
e
co
mmu
n
i
ca
ti
on: A su
rvey
.
J
ourn
a
l
of C
h
o
ngq
ing
U
n
iver
sity of Posts
and
T
e
l
e
co
mmu
n
ic
ations
(Natural Sc
ien
c
e Editio
n)
. 20
15; 27(4): 4
41-
452.
[3]
Y Yifei, Z
L
o
n
g
m
ing. A
ppl
icati
on sc
enar
ios
a
nd E
nab
lin
g T
e
chno
log
i
es
of 5
G.
Communic
a
tions.
2
014;
11(1
1
): 69-7
9
.
[4]
IMT
-
2020(5G)
Promotio
n Group. 5G vision
a
nd dem
an
d
.
20
14; 5.
[5]
You
Xia
ohu,
P
an Z
h
i
w
e
n
, Ga
o
Xi
qi, et
al. T
he
5G mo
bil
e
communic
a
tio
n
:
the d
e
ve
lop
m
ent tren
ds
and its emer
gi
ng ke
y tech
niq
ues.
SCIENCE
CHINA Info
rmation Sci
enc
es.
2014: 4
4
(5): 5
51-5
63.
[6]
Yuan Z
h
ifen
g, Yu Guan
gh
ui
, Li W
e
i
x
i
ng.
Mu
lti-User S
h
ared Acc
e
ss for 5G W
i
rel
e
s
s
Net
w
orks.
Network Technology
. 201
5; (5): 28-31.
[7]
H Niko
p
o
u
r, H
Bali
gh.
Spars
e
Co
de Mu
ltip
le Access
. IEEE International Sy
mposium on PIMRC.
201
3: 332-
336.
[8]
Kang
Shaoli, Dai Xiao
ming, Ren Bin. Pattern Di
vision Multiple Acc
e
ss
fo
r 5G Wireles
s
Net
w
orks
.
201
5; (5): 43-4
7
.
[9]
T
ao Yunzhe
ng
, Liu L
ong,
Li
u Sha
ng, et
al. A
surve
y
: Severa
l
techn
o
lo
gies of
no
n-ortho
g
o
n
a
l
transmission for 5G.
Commu
n
i
c
ations.
20
15;
12(1
0
): 1-15.
[10]
W
ang Bich
ai, W
ang Kun, L
u
Z
hao Hua, et
al.
Co
mp
ariso
n
study of Non-
Orthogon
alMu
l
t
iple Access
sche
m
es for
5G.
IEEE Internati
o
n
a
l S
y
mp
osi
u
m on
Broadb
an
d Multimed
ia S
y
stems a
n
d
Broadc
astin
g
. 201
5: 1-5.
[11]
F
a
rzad Moaz
zami. Succes
s
ive Co-C
ha
n
nel Interfer
en
ce Canc
ell
a
tio
n
.
Indon
esia
n
Journa
l of
Electrical E
ngi
neer
ing a
nd Inf
o
rmatics (IJEEI)
. 2015; 3(1): 3
4
-38.
[12]
Ahmed
El-sa
y
e
d
Ei-M
ahd
y,
M
age
d A
h
med.
L
o
w
Co
mp
le
xit
y
Selectiv
e A
dap
tive Mu
lticarri
er
DSC
D
MA
Receiv
er.
T
E
LKOMNIKA T
e
leco
mmu
n
icati
o
n Co
mp
uting E
l
ectron
ics and
Contro
l.
201
3; 2: 321-3
30.
[13]
W
u
Kai, San
g
Lin, W
a
n
g
He,
et al. Detecti
o
n
Algorithm for
V-BLAST
Sy
st
ems
w
i
th N
o
vel
Interferenc
e
Canc
ellation T
e
chnique. IEEE Vehicular
T
e
chnology
Conf
er
enc
e. 2009: 1-5.
[14]
Luo Z
h
en
gd
on
g, Liu
Si
ya
ng,
Z
hao Mi
ng,
et al.
Gen
e
ral
i
z
e
d P
a
ral
l
el
Interferenc
e C
ance
llati
o
n
Algorit
hm
for V
-
BLAST System
s
. IEEE Inter
national Co
nference on
Com
m
unic
a
tion. 2006;
7: 3207-
321
2.
Evaluation Warning : The document was created with Spire.PDF for Python.