www.seipub.

org/ijc International J ournal of Communications (IJ C) Volume 2 Issue 1, March 2013

12
ANovelMAIReductionMethodUsing
LongerMsequencesforMultiuserDetection
inDS/CDMACommunicationSystems
AbdulhamidZahedi
*1
,HamidrezaBakhshi
2

DepartmentofEngineering,ScienceandResearchBranchofTehranIslamicAzadUniversity,Departmentof
EngineeringofShahedUniversity
HesarakTehran1477893855Iran,PersiangulffreewayTehran3319118651Iran

*1
[email protected];
2
[email protected]

Abstract
In this paper, we consider the problem of Multiuser
Detection(MUD)inDirectSequence/CodeDivisionMultiple
Access (DS/CDMA) communication systems. The optimum
detectorforMUDistheMaximumLikelihood(ML)detector,
but its complexity is very high and involves an exhaustive
search to reach the best fitness of the transmitted and
received data. Thus, there has been considerable interest in
suboptimal multiuser detectors with less complexity and
reasonable performance. If we look to the problem from an
optimization viewpoint, what makes our work difficult is
highinteractionandEpistasisamongvariableswhichshould
beoptimized.Weproposetoworkontheproblemandtryto
reduce its complexity instead of using more complex and
powerful optimization algorithms. To do so, we analyze
problem carefully and propose a new method to overcome
Epistasis. In this paper, by using sign detector and longer
codelengthsinDS/CDMAsystem,asimpledetectorwillbe
proposed. This method is compared with ML and two
modelfree optimization methods: Particle Swarm
Optimization (PSO) and Ant Colony Optimization (ACO)
algorithms which have been used for MUD in DS/CDMA.
Simulationresultsshowourmethoddramaticefficiency.
Keywords
DirectSequence/CodeDivisionMultipleAccess(DS/CDMA);
MultiuserDetection;Epistasis;Msequencecodelength;Sign
Detector
I nt r oduc t i on
InaDS/CDMAsystem,thereceiverisaMatchedFilter
Bank (MFB), which comprises the conventional
detector(signdetector).Thistypeofreceiverisunable
to optimally recover the signal, when the channel is
contaminated by additive white Gaussian noise
(AWGN), and suffers from flat or frequency selective
fading; because the DS/CDMA signal is affected by
Multiple Access Interference (MAI) and also by the
NearFarRatio(NFR)[1].Infact,thesignaturesignals
of different users are not completely orthogonal to
each other, and cross correlation among these signals
results in multiple access interference. Therefore, the
conventional matched filter detector, as in single user
communication, is no longer effective and causes
manyproblems[2].In1986,Verduin[3]proposedthe
optimum multiuser detector (OMUD) which consists
of a bank of matched filters followed by a maximum
likelihood sequence estimator (MLSE). The MLSE
detectorgeneratesamaximumlikelihoodsequence,

b
,
which is associated with the transmitted sequence, as
presented in Fig. 1 [1]. The vector
b
is estimated in
order to maximize the sequence transmission
probability given that
( ) r t
is received; where
( ) r t
is
extended for all messages, considering all the
transmitted messages with the same transmission
probability [1]. The OMUD has a computational
complexity which grows exponentially with the
number of users. Thus, since CDMA systems could
potentiallyhavealargenumberofusers,theOMUDis
impractical to implement for them. Therefore, many
researches have focused on suboptimum detectors
with less complexity and a performance which is
almostashighastheOMUD.Alternativedetectorsfor
OMUDincludetheDecorrelatorproposedbyVerduin
[3]andtheMMSEdetectorrecommendedbyPoorand
Verdu in [4]. These algorithms have reasonable
computational complexity, and their performance is
comparable to that of the optimum receiver, but they
yield a degraded communication system in the sense
ofBER[5].
According to the problem of ML detector, many
methodsforsuboptimaldetectionhavebeenproposed.
Someheuristicmethodshavebeendeveloped,suchas
genetic algorithm (GA). The first GAbased multiuser
International J ournal of Communications (IJ C) Volume 2 Issue 1, March 2013 www.seipub.org/ijc
13
detector (GAMUD) was proposed by Juntti et al. [6]
which assumed a synchronous CDMA system model
communicating over an AWGN channel. After that in
[7, 8] the new approach of GA is proposed. Tabu
search algorithm [9] and simulated annealing
algorithm(SAA)[10,11]aretheothernewapproaches
ofMultiuserdetectors.
In fact, all the approaches for MUD are classified into
twoglobalapproaches:thefirstisthepropositionofa
new suboptimal Multiuser detector with tradeoff
between performance and complexity such as MUD
based on Particle Swarm Optimization (PSO) and
wavelet transform [1217] and Ant Colonybased
MUD in [1821]. The second approach is the reducing
ofMAIandapplyingsimpleconventionaldetector.In
other words, instead of using more complex methods
for multiuser detection, we try to reduce Multiple
Access Interference in DS/CDMA system and then
apply simple detector such as conventional detector.
Forthispurpose,wemustintroduceEpistasis.
Epistasis is a phenomenon where the effects of one
geneareinfluencedbyoneorseveralothergenes.The
word Epistasis was used by William Batson [22] in
genetic science to describe a gene that affects the
operationofothergenes.

FIG.1BASEBANDDS/CDMABLOCKDIAGRAM,RECEIVER
WITHMUD[1].
In optimization literature, Epistasis is used with
similar definition: the interaction between
optimizationvariables[23].HighlyEpistaticproblems
areverymoredifficulttobeoptimized.Infactinevery
nontrivial optimization problem, there should be at
leastalittledegreeofEpistasis;otherwise,theproblem
will be trivial [24]. Since in such cases, the Nvariable
nonEpistatic function can be divided into N separate
functions each with one variable. In our optimization
problem, MAI is the cause of Epistasis.So to designa
suitable detector for DS/CDMA, we try to reduce
Epistasis (here, MAI) and use conventional detector
ratherthanusingmorecomplexMultiuserdetectors.
MAI is the result of correlation between signature
signalsofusers.Correlationcouldbeshowninsquare
matrix form in which elements on main diagonal are
autocorrelations and elements outside main diagonal
arecrosscorrelationsbetweenthesignaturesignalsof
users. For Msequence codes if code with higher
length is used, it will be claimed that autocorrelation
elements will remain constant but cross correlation
elements decrease proportional to inverse of code
length [25], i.e. longer Msequence codes, less MAI. It
is important to note that long codes have some
advantages such as interference randomization and
improving the system capacity. Based on the above,
our analysis focuses on DS/CDMA systems with long
spreading codes. Our proposed method is compared
to the ML detector, Matched Filter detector, Ant
Colony and PSObased detectors, and we show that
theperformanceoftheproposedmethodisbetterwith
lesscomplexity.
The remainder of this paper is organized as follows:
Section 2 describes our asynchronous CDMA system
and section 3 highlights the simple method used to
implement our proposed detector. The simulation
results are presented in Section 4, while Section 5
provides comparison of the complexities associated
with our algorithm and several effective methods.
Finally,someconclusionsaredawninSection6.
Syst em Model
InaDS/CDMAsystemwithbinaryphaseshiftkeying
modulation (BPSK) shared by k asynchronous users,
asillustratedinFig.1,thekthusertransmittedsignal
isgivenby[1]:
( )
( )
( ) ( ) 2 cos .
i
k k k k b c
i
x t p b s t iT t e =

(1)
Where
k
p
representsthekthusertransmittedpower;
( ) i
k
b
istheithBPSKsymbolwithperiod
b
T
;
c
e
isthe
carrier frequency and
( )
k
s t
corresponds to the
spreadingsequencedefinedintheinterval
| ) 0,
b
T
:
( ) ( )
1
,
0
; 0
N
k c k n b
n
s t p t nT c t T

=
= s s

(2)
www.seipub.org/ijc International J ournal of Communications (IJ C) Volume 2 Issue 1, March 2013
14
Where
{ }
,
1, 1
k n
c e +
is the nth chip of the sequence
with length N used by the kth user;
c
T
is the chip
periodandthespreadspectrumprocessinggain,
b
c
T
T
is
equal to N ; the pulse shaping
( ) p t
is assumed
rectangular with unitary amplitude in the interval
| ) 0,
c
T
andzerooutside.
Assuming a frame with I bits for each user
propagating over L independent slow Rayleigh
fadingpaths,thebasebandreceivedsignalinthebase
stationis[1]:
( )
( )
( )
( )
( ) ( )
1
0 1 1
.
I K L
i i
k k k k k
i k l
r t A b s t h t w t t

= = =
= - +

(3)
Where K is the number of active users,
| ) 0,
b
t T e
.
The amplitude
k
A
is assumed constant for all I
transmitted bits,
{ } 1, 1
k
b e +
is the transmitted
information bit,
k
s
denotes a copy of the signature
sequence assigned to the kth user and
k
t
represents
the random delay associated to the kth user; the
complexlowpassimpulseresponseofthechannelfor
thekthuserovertheithbitintervalcanbewrittenas
[1]:
( )
( )
( )
( )
, ,
1
.
L
i i
k k l k l
l
h t a t o
=
=

(4)
Where
, k l

is the propagation delay and

( )
,
i
k l
a
is the
complex channel coefficient whose amplitude has
Rayleigh distribution, and its phase is uniformly
distributed over
| | 0,2t
. Finally, w(t) represents the
AWGNwithbilateralpowerdensityequalto
0
2
N
.
Usingvectorialnotation,Equation.(3)canbestatedas
[1]:
( ) ( )
( )
( )
1
( )
0
.
I
i T i
b
i
r t s t iT Aa b w t

=
= +

(5)
Where
| |
1 2
, ,...,
k
A diag A I A I A I =
is the diagonal
matrix for the users amplitude including the path
losses and shadowing effects, and
L L
I

is the identity
matrix with a dimension equal to L ; s is the vector
of users signature sequence, and a is the diagonal
channelgainmatrixas:
( ) ( ) ( ) ( ) ( ) ( )
1,1 1, 2,1 2, ,
,..., , ,..., ,...,
i i i i i i
L L K L
a diag a a a a a ( =


(6)
Andthedatavectorisgivenby:
( ) ( ) ( ) ( )
1 2
, ,...,
T
i i i i
K
b b b b
(
=

(7)
representingthe1Lkthuserbitvector.
If we use the conventional Rake receiver which
consists of a bank of KL filters matched to the users
signaturesequence[2], then the output for the ith bit
intervalcanbeexpressedas[1]:
( )
( ) ( )
( ) ( ) ( ) ( )
, , , , ,
.
i i i i i
k l k b k l k b k l k k l k l
y r t s t iT dt A T b I n t
+

= = + +
}
(8)
Wherethefirsttermcorrespondstothedesiredsignal,
the second refers to the MAI over the th multipath
componentofthekthuser,andthelastrepresentsthe
filtered AWGN. It should be noted that the auto
interference term is neglected for simplicity. MAI is
the result of cross correlation between signature
signalsofusersasexpressed[1]:
( ) ( ) ( )
,
0
,
b
T
j k j k b
R i s t s t iT dt t t = + +
}
(9)
IfweusetheRakereceiver,weneedtoestimatesome
parameterssuchaschannelcoefficientsanddelay
( ) t
.
When the number of users increases, the significance
of interference rises and owing to this fact, the
conventional detector performance in MUD is
degraded. As mentioned in the previous section, we
use the optimum detector to solve this problem. The
optimum detector is the maximum likelihood
sequencedetectorthatselectsthemostlikelysequence
of transmitted bits given by the observations at the
receiver. In this context, the Kuser, Lpaths, Iframe
andasynchronouschannelschemecanbeviewedasa
KLIuser synchronous channel scheme, and then the
KLIuservectorBcanbewrittenas[1]:
International J ournal of Communications (IJ C) Volume 2 Issue 1, March 2013 www.seipub.org/ijc
15
( ) ( ) ( ) 0 1 1
, ,...,
T T T
T
I
B b b b

(
=
(

(10)
Basedon[27],Verduprovedthatinordertoselectthe
maximum likelihood sequence B, we must maximize
theloglikelihoodfunction(LLF):
( ) { }
2Re .
T H T H
f B B a Ay B aARAa B =
(11)
Where R is the cross correlation matrix and y is the
outputvectoras[1]:
(0) (1) ( 1)
, ,...,
T T T
T
I
y y y y

(
=

(12)
Neglecting the channel effect, we can state (11) in a
simpleformas[1]:
( ) .
T T
f B B r B RB =
(13)
Where r is the received signal and B is the
transmitted sequence to be guessed and we neglect
otherparametersin(11)duetothechanneleffect.The
completeframewiththeestimatedtransmittedbitsfor
all K users can be obtained through optimization of
(13),resulting[1]:
( )
{ } 1, 1
max

arg
IK
B
f B
B
e +
(


=
`

)
(14)
The OMUD attempts to find the best vector of data
bits but because of high complexity and unfeasible
implementation, it is an inefficient method for
multiuser detection. Because of the high
dimensionality of OMUD, all suboptimal algorithms
try to find a solution following an objective function
whichisabletoimprovetheperformanceofmultiuser
detection.Theseattemptstrytoreducethecomplexity
of OMUD and maximize the DS/CDMA mean
performance.Mosteffortsconcentrateonapproaching
theperformanceofMLalgorithmwithlesscomplexity
and reliableapplicabilityalong with the least possible
error.Inthenextsection,weproposeouralgorithmto
achieve this goal and compare our algorithm to other
efficientalgorithmsavailableintheliterature.
Pr oposed Al gor i t hm
SignDetector
In conventional single user detection, the receiver for
eachuserconsistsofademodulatorthatcorrelatesthe
receivedsignalwiththesignaturesequenceoftheuser
and passes the correlator output to the sign detector,
whichmakesadecisionbasedonthesinglecorrelator
output.Thusthesigndetectorneglectsthepresenceof
theotherusersofthechannelorequivalentlyassumes
that the aggregate noise plus interference is white or
Gaussian. Clearly if the signature sequences are
orthogonal,theinterferencefromotherusersvanishes
and the conventional signal detector is optimum.
Because it is not possible to design signature
sequencesforanypairofusersthatareorthogonalfor
all time offsets, thus the conventional detector is
vulnerabletointerferencefromotherusers[2].
Now we should choose between two options. Using
more complex and powerful detectors to overcome
MAI and give a suboptimum detector near ML
detector, or try to eliminate Barriers mentioned above
insigndetectormethod.
EpistasisandMAI
In genetic science, Epistasis is a phenomenon where
theeffectsofonegeneareinfluencedbyoneorseveral
other genes. This concept exists in optimization
problems too where the values of one optimization
variables cannot be labeled (as good, bad and so on)
without considering the status of other variables. For
example consider the optimization test functions in
tables 1 and 2. These functions are some of the most
famoustestfunctionswhicharecommonlyusedinthe
literature [2731]. In table 1 you can see some non
epistatic problems. In these problems you can
optimize the objective function for each variable
separately, while in the problems such as the ones in
table2,theimportantmatteristhearrangementofthe
variables and their values beside each other. In such
problems there is no superposition and the whole is
something more than sum of the individuals. That
extra comes from the information existing in the
structure.
TABLE1FUNCTIONSWITHZERODEGREEOFEPISTASIS
Function Formula
1
F

2
1
N
n
n
x
=

2
F

( ) ( ) sin 4 1.1 sin 2 x x y y +

3
F

( )
2
1
10 10cos 2
N
n n
n
N x x t
=
( +

www.seipub.org/ijc International J ournal of Communications (IJ C) Volume 2 Issue 1, March 2013

16
TABLE2EPISTATICFUNCTIONS
Function Formula

4
F

| |
{ }
2
2
2
1
1
100 1
N
n n n
n
x x x
+
=
+ (

5
F

( )
( )
2 2 2
2 2
sin 0.5
0.5
1 0.1
x y
x y
+
+
+ +

Next,letustakeacloserlookattheLLF.Considering
Equation (13), the objective function is comprised of
two terms:
T
B r
and
T
B RB
. For maximizing the
T
B r

part, considering the fact that each element of B can

only take the values of either 1 or 1, it is enough for
eachelementof B tobeeither1or1accordingtothe
sign of corresponding element in r . Since the
T
B RB

term is a square matrix, if the nondiagonal elements
were small, it would be practically equal to
T
B B
. In
thiscase,itisinsignificantifthevaluesoftheelements
of B areeither1or1.Themainproblemisthatsome
nondiagonal matrix elements are not considerably
small and cannot be assumed as negligible almost
always. Because the matrix R is symmetric, these
elements add terms of the type
2
T
ij i j
R B B
to the
objective function. In this case if
ij
R
is positive, the
fact that
T
i
B
and
j
B
bearing different sign helps
increasing the value of the objective function, and if
ij
R
is negative, their similar signs help this fact. In
other words, the problem variables are no longer
independentandtheirinteractionisinfluential.
In the first glance it seems that we should try to find
theoptimizationmethodswhichhavepotentialability
toovercomethisproblemtosomeextentasin[13,15].
ButwhatifwecoulddecreasetheEpistasisandmake
the problem trivial. If we do this we can solve the
problemwithsigndetectoreasily.
UsingLongerMSequenceCodes
If there is not much limitation on band width in
communication system, cross correlation between
codes can be decreased using longer Msequence
codes. In other words the probability of user
interference will be reduced and this means that
elementsthatarenotonthemaindiagonalofmatrixR,
usingthiswork,aredecreasednoticeablybutthemain
diagonalelementsarefixed.
For Msequence codes if codes with higher length are
used, cross correlation will be reduced according to
(15) proportional to 1/N, [25] i.e. longer Msequence
codes,lessEpistasis.
1
1 ij
if i j
R
if i j
N
=

=

=

(15)
So by using longer codes, Epistasis would be
decreased and sign detector could be applied for
detecting the transmitted signal with minimum error.
This property is also verified for Gold codes or other
codes that used in some communication systems for
spreading signals. Therefore by using longer codes,
thecapacityofsystemisincreasedandtheinterference
amongusersisdecreasedandconsequentlythesimple
andlowcomplexitydetectorscanbeappliedforMUD.
For better presentation of MAI decreasing, one
simulationisdoneandtheaverageofcrosscorrelation
of users in different code lengths is evaluated and
depicted in Fig. 2. In this figure, it is showed that
increasing the code lengths reduced the cross
correlation mean value exponentially rather than
linearly. In fact the relation is followed by
x
y ba

=

that y and x are presentations of cross correlation
mean value and code lengths respectively. From
simulationandusinginterpolation,thevaluesofaand
b are approximated as, b = 10.94 and a = 4. This fact
verifiesourmethodeffectivity.Sincebysmallincrease
in codelength a considerable decrease in MAI is
obtained.

FIG.2CROSSCORRELATIONMEANVALUEVERSUS
DIFFERENTCODELENGTHS,THEEXPONENTIALRELATION.
Numer i c al Resul t s
In this section, the performance of the algorithm
described in Section 3 is compared to Matched filter
receiver(MF),MaximumLikelihoodmethod(ML)and
4 5 6 7 8 9 10 11 12
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
0.045
code length
c
r
o
s
s

c
o
r
r
e
l
a
t
i
o
n

m
e
a
n
International J ournal of Communications (IJ C) Volume 2 Issue 1, March 2013 www.seipub.org/ijc
17
two optimizationbased algorithms, PSO and ACO
considering the BER as the main figure of merit. The
convergence of each algorithm versus optimization
parameters is also provided. It is assumed that the
communication system is asynchronous DS/CDMA
MUD,overslowRayleighfadingAWGNchannel.
The numerical results were obtained based on the
averaging of 1000 simulation runs;and these results
were attained in identical systems and channel
conditionsinordertoprovidefaircomparisontoother
algorithms.Simulationparametersareasfollow:
TABLE3SIMULATIONPARAMETERS
Modulation BPSK
Spreadingcode msequence
Communicationsystem
Uplinkasynchronous
CDMA
Usernumber(K) 20
Channel
AWGNwithslow
Rayleighfading
Pathnumber(L) 4
Pathlossvariance 5dB
The spreading sequencesareselected as pseudonoise
(PN) msequence; in addition the number of active
asynchronous users in the system is K=20; in all
figures, it was assumed that the phases, amplitudes,
channel gains and random delays of all users are
perfectlyknowninthereceiver,anduserspoweris:
( )
2
,
1
1, 1,2,...,
L
i
k l
l
E a for k K
=
(
= =
(

(16)
In these simulations that will be seen, our proposed
method is more effective and better in comparison to
thetwootheralgorithms,ACOandPSO.
In Fig. 3, a comparison between the proposed
algorithm with comprehensive search in the space of
parameters known as ML versus BER of
communication system is proposed. From this figure,
itcanbeseenthatincreasingthelengthofmsequence
code causes the BER decreasing. Especially in code
length equal to 10, the difference between our
algorithm and ML is negligible. So using msequence
with length 10 is suitable and it is used for the
following simulations. Of course notice that codes
withhigherlengthshavemorecomplexityandarenot
suitable for our purpose of simple and low complex
method.ThissimulationisdoneinSNR15dB.
4 5 6 7 8 9 10
10
-3
10
-2
10
-1
code length
B
E
R


Signdetector, different code lengths
ML

FIG.3THEBERCOMPARISONBETWEENMLANDPROPOSED
ALGORITHMWITHDIFFERENTCODELENGTHS(SNR=15dB)
The main parameter of each communication system,
BER, is discussed in Fig. 4. This property of the
proposed algorithm is compared to ACO and PSO
algorithms and also to ML and the worst case as
Matchedfilter.ItisrevealedthatalthoughinlowSNR,
there is no main difference among the methods; the
proposed algorithm in high SNR converges to
optimum MUD and the BER of this method is less
thanthatinACOandPSOmethods.Thisfigureshows
theunsuitabilityofthematchedfilterdetector.
-5 0 5 10 15
10
-4
10
-3
10
-2
10
-1
10
0
SNR(dB)
B
E
R


MF
PSO
ACO
Sign_detector,code length=10
ML

FIG.4BEROFTHEPROPOSEDALGORITHMINCOMPARISON
TOOTHERMETHODSVERSUSSNR.
In Fig. 5, the effect of increasing the number of users
on BER is analyzed for the classic algorithms and
optimizationbasedtechniquessuchasACOandPSO.
Onthebasisofthisfigure,itcanbeseenthattheACO
and the proposed algorithm almost work similarly,
and BER of these two algorithms does not
considerablyincreasewithincreasingusernumbers.It
can also be observed that the performance of the
www.seipub.org/ijc International J ournal of Communications (IJ C) Volume 2 Issue 1, March 2013
18
classicmatchedfiltermethodisdegradedsubstantially.
ThissimulationisdoneforSNR12dB.

FIG.5BEROFTHEPROPOSEDALGORITHMVERSUSNUMBER
OFUSERSINCOMPARISONTOOTHERMETHODS.SNRIS
FIXEDAT12DB.
In Fig. 6, the proposed algorithm with different code
lengths is implemented and the numerical results are
depicted. It is observed that with the code length 10,
the algorithm converges to optimum MUD in high
SNRs. Of course, for the other code lengths less than
10 such as 8 and 6, the error is small and the
convergence is acceptable to some extent. As seen
fromthisfigure,theproposedalgorithmoperatesvery
well, and the convergence of this case is quite
interesting.
-5 0 5 10 15
10
-4
10
-3
10
-2
10
-1
10
0
SNR(dB)
B
E
R


MF
Signdetector, code length=6
Signdetector, code length=8
Sign_detector,code length=10
ML

FIG.6BEROFTHEPROPOSEDALGORITHMVERSUSSNRFOR
DIFFERENTCODELENGTHS,COMPAREDTOMLANDMF.
Comput at i onal Compl ex i t y
A common form in order to compare algorithms
complexities can be done through the O notation,
which means the order of magnitude associated with
the algorithm complexity. But comparing algorithms
only withO can be insufficient. In order to express
the complexity of the analyzed algorithms, it is
essential to determine which instructions are carried
out and how many times they are processed [1].
Suppose the number of transmitted bits is I and the
number of users is K . For OMUD, the number of
operationsincreasesexponentiallywiththenumberof
users, i.e.
2
KI
. For our proposed algorithm, the code
length and the size of correlation matrix of m
sequence codes are the most important factors of
calculation.Afterthat,thesigndetectorisverysimple
and has no complexity. The complexity of correlation
matrix by code length 10 is fixed and it is calculated
andstoredbeforereceiverindatabaseofthesystemto
use when needed. Fig. 7 shows that the complexity
associated with our proposed algorithm is much less
than ML and comparable to matched filter and ACO
detector.

FIG.7COMPUTATIONORDEROFTHEPROPOSEDMETHOD
VERSUSTHENUMBEROFUSERS,COMPARISONTOACO,ML
ANDMATCHEDFILTERMETHOD.
Conc l usi ons
In this paper, multiuser detection based on a new
algorithm was implemented, and through
presentation of msequence codes and sign detector,
the desirable method was achieved. When compared
to suboptimal algorithms such as ACO and PSO, the
newalgorithmshowsbetterperformance,andinsome
cases, approaches the optimal ML algorithm with
minimum error. This algorithm is quite efficient, and
showsacceptableperformanceincodelengthequalto
10. Of course it should be noted that code lengths
more than 10 has more complexity and more
calculation time and codes with shorter length have
notgoodperformanceasthe10lengthone.
5 6 7 8 9 10 11 12 13 14 15
10
-3
10
-2
10
-1
Number of users
B
E
R


MF
PSO
ACO
Sign_detector,code length=10
ML
5 10 15 20 25 30
10
2
10
4
10
6
10
8
10
10
10
12
Number of Users
C
o
m
p
u
t
a
t
i
o
n

O
r
d
e
r


MF
ACO
Signdetector,code length=10
ML
International J ournal of Communications (IJ C) Volume 2 Issue 1, March 2013 www.seipub.org/ijc
19
ACKNOWLEDGMENT
Authors want to acknowledge ITRC organization of
Iranbecauseofitssupportandassistance.
REFERENCES
C. Ergun and K. Hacioglu, Multiuser Detection Using a
GeneticAlgorithminCDMACommunicationssystems,
IEEE Trans. Communications, vol. 48, pp. 522561, Aug.
2000.
D. H. Kim, A. Abraham, J. H. Cho , A hybrid genetic
algorithm and bacterial foraging approach for global
optimization, journal of Information Sciences, vol. 177,
pp.39183937,April2007.
E. Alba, F. Luna, A.J. Nebro, J.M. Troya, Parallel
heterogeneous genetic algorithms for continuous
optimization,journalofParallelComputing,vol.30,pp.
699719,May,2004
Fernando Ciriaco, Tauk Abrao, Paul Jean E. Jeszensky,
DS/CDMA Multiuser Detection with Evolutionary
AlgorithmsJournalofUniversalComputerScience,vol.
12,no.4,pp.450480,2006.
H. H. ElMora, A. U. Sheikh, A. Zerguine, Application of
particle swarm optimization algorithm to multiuser
detection in CDMA, in: Proc. IEEE 16th PIMRC, vol. 4,
pp.25222526,Sep.2005.
J. Andre, P. Siarry, T. Dognon, An improvement of the
standard genetic algorithm fighting premature
convergence in continuous optimization, Advances in
EngineeringSoftware,vol.32,pp.4960,Dec.2000.
J.G.Proakis,DigitalCommunications,3rded.,McGrawHill,
Singapore,1995.
Jose Carlos Marinello Filho, Reginaldo Nunes de Souza,
TaufikAbraoAntcolonyinputparametersoptimization
for multiuser detection in DS/CDMA systems Expert
Systems with Applications, vol. 39, Issue 17, pp. 12876
12884.Dec.2012.
L. D. Davis, K. De Jong, M. D. Vose, L. D. Whitley,
Evolutionary Algorithms, SpringerVerlag New York,
Inc.,1999.
L. Hongwu , An ACO Based Multiuser Detection for
ReceiveDiversity Aided STBC Systems, ISECS
International Colloquium on Computing,
Communication,Control,andManagement,CCCM2009,
pp.250253,Aug.2009.
Li Yanping, Zong Hengshan, Meng Deqian, Chang
XiaomingAdaptiveIATCNNMultiuserDetectorBased
on Simulated Annealing of Optimization journal of
EnergyProcedia,vol.17,PartA,pp.588594,2012.
L. Lin, A Novel Genetic Multiuser Receiver Based on
Wavelet Transform and Hamming Sphere Solution
Space,4thIEEEinternationalconferenceoncircuitsand
systems for communications, ICCSC 2008, pp. 260264,
May2008.
L. xiazhi, f. dawei, z. tianshuang, multiuser detection of
DSCDMA based on PSOICA, IEEE international
conference on image and signal processing, pp. 2490
2493,Oct.2011.
M. Bessaou, P. Siarry, A new tool in electrostatics using a
reallycoded multipopulation genetic algorithm tuned
through analytical test problems, Advances in
EngineeringSoftware,vol.32,pp.363374,May2001.
M. Jiang , Ch. Li , D. Yuan , M. A. Lagunas , Multiuser
Detection Based on Wavelet Packet Modulation and
Artificial Fish Swarm Algorithm, IET conference on
wireless, mobile and sensor network, (CCWMSN07),
china,pp.117120,2007.
M. J. Juntti, T. Schlosser, J. O. Lilleberg, Genetic algorithms
for multiuser detection in synchronous cdma, in proc.
IEEE International Symposium on Information Theory
(ISIT97),Ulm,Germany,June29July4,1997,p.492.
M. L. Steinberg and S. D. Cosloy, Dictionary of
Biotechnologyand Genetic Engineering, 3rd Edition, the
FactsonFile,2006.
Nan Zhao, Xianwang Lv, Zhilu Wu A hybrid ant colony
optimization algorithm for optimal multiuser detection
in DSUWB system Expert Systems with Applications,
vol.39,Issue5,pp.52795285,April2012.
R. L. Haupt and S. E. Haupt, Practical Genetic Algorithms,
2ndEdition,JohnWiley&SonsInc,2004.
R. Lupas, S. Verdu, Linear multiuser detectors for
synchronous code division multipleaccess channels,
IEEE Trans. Inform. Theory, vol. 35, pp. 123136, Jan.
1989.
www.seipub.org/ijc International J ournal of Communications (IJ C) Volume 2 Issue 1, March 2013
20
R. Peterson, R. Ziemer, D. Borth, Introduction to Spread
Spectrum Communications, Englewood Cliffs, NJ,
PrenticeHall,1996.
S. Verdu, Minimum probability of error for asynchronous
Gaussian multiple access channels, IEEE Trans. Inform.
Theory,vol.32,pp.8596,Jan.1986.
S. Verdu, Multiuser Detection, Cambridge Univ.Press,
Cambridge,U.K.1998.
Tang. P. Y., Li. Z. H and Huang. S. J, Multiuser Detector
BasedonGeneticAlgorithmandTabuSearch,Journalof
UniversityofElectronicScienceandTechnologyofChina,
Chengdu,vol.5,pp.499509,2004.
V. H. Poor, S. Verdu, Probability of error in MMSE
multiuserdetection,IEEETrans.Inform.Theory,vol.43,
pp.5060,May1997.
X. Chong, R. G. Maunder, L. L. Yang, L. Hanzo, Near
Optimum Multiuser Detectors Using SoftOutput Ant
ColonyOptimization for the DSCDMA Uplink, IEEE
SignalProcessingLetters,vol.16,pp.137140,Feb.2009.
Y. T. Kao, E. Zahara, A hybrid genetic algorithm and
particle swarm optimization for multimodal functions,
AppliedSoftComputing,vol.8,pp.849857,Mar.2008.
Y. YanFei, Zh. YuanPing, Immuneendocrine genetic
algorithmformultiuserdetectorproblem,International
Conference on Computer Science and Software
Engineering,pp.447450,Dec.2008
Ye. Y. Y and Hu. Y. L, The Application of Simulated
Annealing Genetic Algorithm for Multiuser Detection,
Transactions of Shenyang Ligong University, Shenyang,
vol.6,pp.2325,2007.
Y. Zhao, J. Zheng, Multiuser detection employing particle
swarm optimization in spacetime CDMA systems, in:
Proc.ISCIT2005,pp.940942,Oct.2005.
Z. Guo, Y. Xiao, M. H. Lee, Multiuser detection based on
particle swarm optimization algorithm over multipath
fading channels, IEICE Trans. Communications. E90B,
pp.421424,Feb.2007.

Abdulhamid Zahedi was born in

Kermanshah, Iran, in 1983. He received the
Bs.C. of electrical engineering in Amirkabir
university of technology in Iran in 2005 and
M.Sc. degree in shahed university in Iran in
communicationengineeringin2008.
Currently he is Ph.D. candidate in electrical engineering in
Science and Research branch of Tehran, Islamic Azad
university,Iran.HisinterestfieldsofresearchareMultiuser
Detection andwireless telecommunication,specially CDMA
andMIMOOFDMcommunications.
Hamidreza Bakhshi was born in Tehran,
Iran on April 25, 1971. He received the B.Sc.
degreeinelectricalengineeringfromTehran
University,Iranin1992,theM.Sc.andPh.D.
degree in Electrical Engineering from
Tarbiat Modarres University, Iran in 1995
and 2001, respectively. Since 2010, he has
been an Associate Professor of Electrical Engineering at
ShahedUniversity,Tehran,Iran.
His research interests include wireless communications,
multiuserdetection,andsmartantennas.