Read without ads and support Scribd by becoming a Scribd Premium Reader.
 
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 5, August 2010
Blind Multiuser Detection in SDMA-aided MIMOOFDM Systems by FastICA Algorithm
Yu Du, Kang Yen, Yechang Fang and Nansong Wu
 
Department of Electrical and Computer EngineeringFlorida International UniversityMiami, Florida, United StatesE-mail: {dyu002, yenk, yfang003, nwu001}@fiu.edu
 Abstract
—The techniques of MIMO (Multiple Input and MultipleOutput) OFDM (Orthogonal Frequency Division Multiplexing)systems have been adopted by IEEE 802.11n wirelesscommunication standard (Wi-Fi). Multiuser Detection by blindsignal separation (BSS) techniques in MIMO OFDM systemsrecently attracts researchers’ interests by their variousadvantages. The performance of a MIMO OFDM system mainlydepends on reliability and effectiveness of the system, and themeasurement of reliability is bit error ratio (BER). On thepremise of effectiveness, a new blind multiuser detectionapproach based on FastICA (fast fixed-point algorithm forindependent component analysis) is proposed for a SDMA-aided(Spatial Division Multiple Access) MIMO OFDM system withadditive white Gaussian noise (AWGN) for the target of betterreliability. Compared with the usually-used spatial-multiplexing-based MIMO OFDM system, the proposed method offers betterBER performance. Simulation results confirm the betterreliability of the proposed approach.
 Keywords- Blind Multiuser Detection; SDMA; MIMO OFDM; FastICA
I.
 
I
NTRODUCTION
 The techniques used in MIMO (Multiple Input and MultipleOutput) OFDM (Orthogonal Frequency Division Multiplexing)systems have been adopted by IEEE 802.11n wirelesscommunication standard (Wi-Fi) [1], [2]. For any receiver of MIMO OFDM systems, some priori parameters such ascharacteristics of users and noise, timing information of usersand noise, relative amplitude and training sequences are needed[3]. Blind multiuser detection is a novel approach withoutretransmitting them in MIMO OFDM systems.Blind signal separation algorithms adopted in MIMOOFDM systems usually contain two classifications, which aresecond order statistics (SOS) and higher order statistics (HOS).With regard to SOS methods, a subspace approach with shortaveraging periods has been proposed [4]. Although it achievesa good estimation accuracy, the scheme requires additionalinformation of channel orders for optimizing the ambiguitymatrix, and the computation complexity is not lower when thenumber of subcarriers is larger than 64 [5]. In contrast withSOS, the use of HOS blind signal separation algorithms, whichare also called independent component analysis (ICA) [6],decreases the influence caused by the Gaussian noise of transmission channels in MIMO OFDM systems.The most familiar approach employing HOS in MIMOOFDM systems is Constant Modulus Algorithm [7]. Althoughthis method provides a good solution to inter-symbol-interference (ISI) and inter-user-interference (IUI), it requiresmore receivers than transmitters. Another approach [8] makesuse of tensor based space-time multiplexing and offers sometransmission flexibility, but the problem associated with thismethod is that the transmitters are too complex to beimplemented in practical applications. Well-established HOSmethods in MIMO OFDM systems include JADE (JointApproximate Diagonalization of Eigen-matrices) [5], FastICA[9] and Nature Gradient [10]. The JADE-LSTE (LayeredSpace-time Equalization) approach in [5] applies JADE to allsubcarriers and uses a LSTE approach for solving thepermutation and scaling indeterminacy problems. However, thelimitation of this approach is its complex structure. In [9],FastICA in a MIMO OFDM system is proposed by fractionalsampling in the frequency domain, and the permutation andscaling indeterminacy are solved by a PLL (phase locked loop).The fractional sampling is able to increase the number of received signals and to get better diversity at the receiver.However, this method increases the computation complexity.The Nature Gradient approach [10] is applied to each currentsubcarrier and the previous subcarrier has to be usedsequentially to resolve the permutation and scalingindeterminacy. However, this sequential approach brings someerror propagation. In general, they select similar linear schemes,and because of the higher computation and complexity of nonlinear blind signal separation algorithms, nonlinear schemeshave not been included.To realize a MIMO OFDM system, several technologies,which include beamforming, space-time coding [8], SDMA [1],spatial multiplexing [5], [7], [9] and network, have been used.Compared with other approaches, SDMA schemes are able tomaximize the number of users in wireless communicationsystems, and the interference can be drastically reduced byusers’ channel impulse responses [1]. Furthermore, SDMAschemes can let multiple users in the same time or frequency orcode domain to simultaneously share the same bandwidth atdifferent locations by spatial signature, thus the capacity of thesystem is increased.FastICA algorithm is capable of separating linearly mixedcomplex valued source signals with good computationalefficiency [11]. The use of FastICA algorithm can also supply
1http://sites.google.com/site/ijcsis/ISSN 1947-5500
 
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 5, August 2010
enough separating accuracy for mixed complex valued sourcesignals and has most of the advantages of other HOS blindsignal separation algorithms. According to the attraction of FastICA algorithm, it is employed to the SDMA-aided MIMOOFDM system.With enough effectiveness of a MIMO OFDM system, asmaller BER means a better reliability of the system. Theobjective of this paper is to apply FastICA [6], [11], to SDMA-aided MIMO OFDM systems with additive white Gaussiannoise (AWGN) for detecting multiuser signals with better BERperformance. The simulation results show that the proposedapproach offers less BER and than the spatial-multiplexing-based MIMO OFDM systems.This paper is organized as follows. The system model isdefined in Section II. Section III describes the proposedFastICA receiver. The simulation of proposed method ispresented in Section IV, followed by some conclusions inSection V.II.
 
S
YSTEM
M
ODEL
 A.
 
SDMA-aided MIMO Channel
Compared with a single input single output (SISO) systemwhich is equipped with 1 transmitting antenna and 1 receiverantenna, Figure 1 shows a generic uplink SDMA-aided MIMOsystem model, where each of 
l
simultaneous users is equippedwith a single transmitting antennas, while the base station’sreceiver exploits
 p
-element antenna array. The scenario canimprove spectral efficiency of the system, diversity gain andcapacity of the system. Figure 2 [1] illustrates the schematic of an uplink SDMA-aided MIMO OFDM system. In the
th
 subcarrier of the
n
th
OFDM symbol received by the
 p
-elementantenna array, the set of complex signal vector
x
[
n
,
] isconstituted by the superposition of the independent signalsfrom
l
mobile users and corrupted by AWGN. So the receivedcomplex signal can be expressed as [1]
x = Hs + n =
x
+ n
(1)Here the indices [
n
,
] have been omitted for the sake of notational convenience and the vectors
x
,
s,
and
n
are given by
x
= [
 x
1
,
 x
2
, ···,
 x
 p
]
T
(2)
s
= [
s
1
,
s
2
, ···,
s
l
]
T
(3)
n
= [
n
1
,
n
2
, ···,
n
 p
]
T
(4)where T denotes transpose. The frequency domain channeltransfer matrix
H
with
 p
*
l
dimension is given by [1]
H
= [
H
1
,
H
2
, …,
H
l
]
T
(5)where
H
i
(
i
= 1, 2, ···,
l
) is the set of channel transfer functionvectors of the
l
users to each element of the
 p
-element receiver,which is express as [1]
H
i
= [
 H 
i
1
,
 H 
i
2
, …,
 H 
ip
]
T
,
i
= 1, 2, ···,
l
. (6)There assumptions made in this model are both thecomplex source signal
s
i
and AWGN signal
n
 p
that have zeromean and a variance of 
2
l
δ  
, and the frequency channel transferfunctions
 H 
ij
(
i
= 1, 2, ···,
l
and
 j
=1, 2, ···,
 p
) are independentand stationary. The complex Gaussian distributed processes of the frequency channel transfer functions are set to zero-meanand unit variance [1].
Base station with
 p
-element antenna array
......
User 1User 2User
l
......
 
Figure 1. Generic uplink SDMA-aided MIMO system model
s
1
s
2
s
l
 x
1
 x
2
 x
 p
 H 
11
 H 
12
 H 
1
 p
 H 
21
 H 
22
 H 
2
 p
 H 
l
1
 H 
l
2
 H 
lp
......
 p
-elementReceiverAntennaArray
Figure 2. The schematic of uplink SDMA-aided MIMO OFDM system
 B.
 
 DFT-based OFDM systems
OFDM, which is a special multicarrier modulation in thestandard parallel data transmission systems, divides thefrequency domain into orthogonal overlapping subcarriers. Thefrequency spectrums of the subcarriers can guarantee that thesubcarrier signals do not interfere with each other. OFDMsystem alters a frequency selective broadband channel into flatfading narrowband channels so that the system maintains ahigh spectral efficiency.Figure 3 shows a DFT-based OFDM transmission systemand DFT technique can reduce the implementation complexityof OFDM. A sequence of discrete-time BPSK (Binary Phase
2http://sites.google.com/site/ijcsis/ISSN 1947-5500
 
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 5, August 2010
OFDMTransmitterOFDMTransmitterOFDMReceiverOFDMReceiver
Transmitters Receivers
S
1
S
l
s
1
s
l
 x
1
 x
 p
n
1
n
 p
 X 
1
 X 
 p
......
 H 
11
 H 
lp
 H 
1
 p
 H 
l
1
......
BaseStationAntennaArray
 
Figure 4. The SDMA-aided MIMO OFDM System Model
Shift Keying) modulated symbols goes through a serial/parallelconverter, and then each block of symbols,
S
= [
S
(0)
,
S
(1)
,
S
(2)
, ···,
S
(
 N 
-1)
]
T
, is modulated by IDFT. The new vector,
s
= [
s
(0)
,
s
(1)
,
s
(3)
···,
s
(
 N 
-1)
]
T
is demultiplexed into a serial form to generate asingle OFDM symbol. The cyclic prefix (CP) can avoid ISI,and it is inserted for each OFDM symbol at the transmitter andthen is taken away at the receiver. Through the DFTdemodulation at the receiver, the received frequency symbolcan be shown in the following equation.
 X 
(
)
=
 H 
(
)
S
(
)
(7)here the factor
 H 
(
)
is a complex valued scale at
th
frequency.
S/P IDFTP/SCPP/S DFTS/PCPChannel
h
(
)
Input
S
(0)
S
(1)
S
(
 N 
-1)
...
 X 
(0)
 X 
(1)
 X 
(
 N 
-1)
s
(0)
s
(1)
s
(
 N 
-1)
 x
(
 N 
-1)
 x
(1)
 x
(0)
Ouput
.........
 
Figure 3. Block Diagram of DFT-based OFDM systems
Although OFDM scheme increases the implementationcomplex as shown in Figure 3 compared with a traditionalserial modem, the signal in low speed subchannels is lesssensitive to Inter Symbol Interference (ISI). It is worth notingthat the subchannel modulation scheme is not limited to BPSK,and Quadrature Phase Shift Keying (QBSK), 16 QuatratureAmplitude Modulation (16QAM), 64 QAM and relatedmodulation schemes have also been selected in recent years.
C.
 
The SDMA-aided MIMO-OFDM System Model
The combination of SDMA, MIMO and OFDM mayenhance the quality of wireless communication and increasethe throughput of the system. The SDMA-aided MIMOOFDM is illustrated in Figure 4 [1]. The number of antennasdecides the complexity of estimation at the receivers and alarge number of antennas mean a complex multiuser detection.The linear SDMA-aided mixing model at
th
frequency can bedescribed by (8) or in a compact expression shown in (9).
+=
)()(1)()(1)()( 1)(1)(11)()(1
 pllp pl p
nnSS H  H  H  H  X  X 
MMLMOMLM
(8)
X
(
)
=
H
(
)
S
(
)
+
n
(
)
(9)III.
 
T
HE
P
ROPOSED
F
AST
ICA
 
R
ECEIVER
In [5], it is noticeable that the ICA model for MIMOOFDM systems is an instantaneous linear mixture of eachsubcarrier. It is a basic requirement that allows the applicationof ICA to blind multiuser detection in MIMO OFDM systems.In contrast with SOS, HOS blind signal separations have betterBER performance, thus the existing HOS algorithm will beemployed in the proposed system. Among existing HOSalgorithms, Complex FastICA algorithm [6], [11] is adoptedfor the proposed SDMA-aided MIMO OFDM system becauseof its fast convergence, less computation complexity, and high-quality BER performance.
3http://sites.google.com/site/ijcsis/ISSN 1947-5500
Search History:
Searching...
Result 00 of 00
00 results for result for
  • p.
  • More From This User

    Notes
    Load more