You are on page 1of 5

IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 7, NO.

12, DECEMBER 2008 4849

Efficient Spatial Covariance Estimation for Asynchronous Co-channel


Interference Suppression in MIMO-OFDM Systems
Qiang Li, Jing Zhu, Qinghua Li, and C. N. Georghiades

Abstract—We present algorithms to suppress the asynchronous Access / Collision Avoidance). Thomas et al. [8] showed
co-channel interference (CCI) in MIMO OFDM systems. The that the conventional frequency domain CCI cancelation that
key challenge is that the cyclic prefix of the interference signal estimates both the intended and interfering channels could
does not line up with that of the intended signal due to the
asynchronous transmission in WLAN. Therefore, the orthogo- not work effectively. This is because the asynchronousness
nality across the tones of the interference signal is destroyed destroys the cyclically padded OFDM symbol structure that
and the conventional frequency domain minimum mean square enables the inter-tone orthogonality. Hence, we adopt a sta-
error (MMSE) cancelation techniques that employ the interfer- tistical approach. We first models the asynchronous CCI as a
ence channel response per tone can not work effectively. To zero-mean, time uncorrelated, and spatially colored stationary
suppress the asynchronous interference, we design an efficient
estimator for the spatial covariance matrix of the interference Gaussian random process and then design an efficient esti-
using Cholesky decomposition and low-pass smoothing. Both a mator for the spatial covariance of the CCI, which utilizes
MMSE and a maximum a posteriori (MAP) receiver are derived the OFDM symbol structure and matrix decomposition tech-
based on the estimated interference statistics. Simulation results niques.
demonstrate the effectivity of our solution.
Index Terms—OFDM, co-channel interference, MMSE, MIMO II. S PATIAL C OVARIANCE E STIMATION FOR
system.
A SYNCHRONOUS I NTERFERENCE
Gaussian distribution matches the asynchronous interfer-
I. I NTRODUCTION
ence statistics very well [11], [12]. The Gaussian approxi-

C O-CHANNEL interferences (CCI) is becoming the lim-


iting factor that dominates the performance in the emerg-
ing high-density WLAN (HD-WLAN)[1], where access points
mation simplifies and eases the receiver design because the
second moment, i.e. covariance, is sufficient to characterize the
interference statistics. We exploit the spatial structure (in the
(APs) are densely deployed. Multiple cells that simultaneously covariance matrix) to suppress the interference. For a MIMO
operate on the same channel cannot be separated far enough in OFDM system with Mt transmitter antennas, Mr receiver
distance and will interfere with each other. Researchers inves- antennas and total K subcarriers, the baseband received signal
tigated CCI suppression extensively since Winters’s seminal on the kth tone is written as
paper [2]. Catreux et al. studied the throughput of interference-
limited MIMO cellular system [3]. Blum [4] investigated the yk (tn ) = Hk xk (tn ) + zk (tn ) , (1)
MIMO capacity under interference with single-user detection.
Dai et al. [5] proposed a multiuser detection technique to where xk (tn ) ∈ CMt ×1 and yk (tn ) ∈ CMr ×1 are the
cancel MIMO CCI for flat fading channels. Li and Sollen- transmitted and received signal; zk (tn ) ∈ CMr ×1 represents
berger [6] designed an adaptive array processing scheme using interference plus noise; Hk ∈ CMr ×Mt is the channel matrix
a MMSE diversity combiner for OFDM modulation and a of kth tone; tn is the time index of the nth OFDM symbol.
time-varying channel. Maltsev et al. [7] proposed an MMSE We design an efficient estimator for the covariance of z(tn )
canceler that estimates the interference covariances per tone on each tone, and then design a Wiener filter suppress inter-
from short training symbols and utilizes the correlation acorss ference. The spatial covariance of z(tn ) on the k th tone can
tones. Besides the physical layer approaches, Zhu et al. [1] be expressed as
proposed a medium access control (MAC) based solution, 1
S−1

which adapts the carrier sensing threshold to control the CCI Rkzz = E{zk (tn )zk (tn )H } = lim {zk (tn )zk (tn )H } ,
S S→∞ t =0
level and whose throughput gain was verified by test-bed n

experiments. where S is the number of training symbols. For WLAN, it


Typically, CCI in WLANs is asynchronous due to the ran- is required to measure the interference statistics over short
dom access protocol, i.e. CSMA/CA (Carrier Sensing Medium duration because different interferers are randomly present
Manuscript received October 28, 2007; revised February 10, 2008 and May with transmission duration comparable to the desired packet.
19, 2008; accepted July 22, 2008. The associate editor coordinating the review The sample averaging covariance estimator is the maximum
of this letter and approving it for publication was Y. Zheng. likelihood (ML) estimator and unbiased. However, it tends
Q. Li and C. N. Georghiades are with the Electrical and Computer
Engineering Department, Texas A&M University (e-mail: {qiangli, georghi- to spread the eigenvalues due to the limited samples. This
ades}@ece.tamu.edu). This work was completed while Q. Li was working as tendency is highly undesirable and often causes a substantial
an intern with Communications Technology Lab, Intel Corporation. degradation in performance. Therefore, we utilize the correla-
J. Zhu and Q. Li are with Communication Technology Lab (CTL), Intel
Corporation (e-mail: {jing.z.zhu, qinghua.li}@intel.com). tion information across OFDM tones to refine the estimation.
Digital Object Identifier 10.1109/T-WC.2008.071201 Moreover, we use the Cholesky decomposition to turn a
1536-1276/08$25.00 
c 2008 IEEE

Authorized licensed use limited to: Univ of Texas at Dallas. Downloaded on September 13, 2009 at 15:27 from IEEE Xplore. Restrictions apply.
4850 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 7, NO. 12, DECEMBER 2008

1x2 SIMO, 64QAM


constrained parameter estimation problem (for positive definite 2

matrix) into an unconstrained one.


0 Chol decomp. + smooth
w/ smoothing
w/o smoothing
A. Temporal Low-Pass Smoothing

Relative Estimation Accuracy F−norm (dB)


S−1
Let R̃kzz  S1 tn =0 {zk (tn )zk (tn )H }, where R̃kzz ∈
−2

CMr ×Mr . The sequence {R̃1zz · · · R̃K zz } fully characterizes the


statistics of the interference. The vector of diagonal entries −4

S̃nn  [R̃1zz [n, n], · · · , R̃K


zz [n, n]]
T
is the estimated power
spectral density (PSD) of interference plus noise received from −6

the nth antenna. Similarly, the vector of off-diagonal sequence


S̃mn  [R̃1zz [m, n], · · · , R̃K T
zz [m, n]] represents the estimated −8
mutual PSD between interference signals received from the
mth and nth antennas. Transforming the auto/mutual PSDs
back to time domain with inverse discrete Fourier transform
−10
8 10 12 14 16 18 20 22 24
SNR (dB)
(IDFT) provides the (cyclic) auto/cross-correlations of the
interference plus noise r̃mn = F−1 S̃mn , for m, n = 1 · · · Mr Fig. 1. Relative estimation accuracy for spatial covariance, defined as:
R̂k −Rk 
and time spacing from 0 to K − 1, where F−1 is the IDFT ξF (R̂kzz , Rkzz ) = zz
Rk
zz F
zz F
where Rkzz is the actual spatial covariance;
matrix. We assume that the transmitted interference signal and AF denotes the Frobenius norm of matrix A. With 4 OFDM symbols silent
time interference measurement. 1 interferer, standard 802.11n channel model
the noise are uncorrelated in time and have a zero mean. D, interleaver and LDPC code are used. The relative estimation accuracies
Because the maximum channel delay is L, the auto/cross- averaged across tones are plotted.
correlation of the received interference plus noise is zero if
the time spacing between the two samples is greater than L, TABLE I
S PATIAL C OVARIANCE E STIMATION A LGORITHM I
i.e. E{zm (τ1 )zn (τ2 )} = rmn (τ1 − τ2 ) = 0 for |τ1 − τ2 | ≥ L,
where zm (τ ) is the time domain interference plus noise from 1. Samples Average Estimation: for k = 1 · · · K,
mth receiver antenna. Therefore, there are at least K − 2L + 1  S−1 k
R̃kzz = S1 H k
tn =0 {z (tn ) z (tn )}.
zeros in the middle part of the auto/corss-correlation vectors 2. Cholesky Decomposition: for k = 1 · · · K, R̃kzz = (Ũk )H · Ũk
 T
r̃mn s. This is referred to as “time domain lowpass” and the 3. Smoothing: for each entry in Ũk , let ũ = Ũ1 [m, n] · · · ŨK [m, n] ,
 1 
limited channel delay implies that the corresponding entries v = P · ũ, then v = Û [m, n] · · · ÛK [m, n] ,
T

of the frequency domain covariance matrices are correlated. k


construct Û from v.
The “lowpass” property can be exploited by filtering co- 4. Reconstructing : reconstruct the estimated covariance,
variance estimates across tones, where the filter matrix P [7] R̂kzz = (Ûk )H · Ûk .
can be pre-computed and stored. It is known that lowpass
filters can smooth temporally correlated signals. Analogically,
the above process can be regarded as a temporal lowpass smoothing the covariance matrices. The Cholesky decompo-
filtering that smooths spectrally correlated signal. However, if sition of R̃kzz is expressed as R̃kzz = (Uk )H Uk , where Uk
we filter each vector S̃mn individually, it is not guaranteed that is a upper triangle matrix; Uk is also called “square-root” of
the smoothed covariance matrix formed by the filtered entries matrix R̃kzz .
for each tone is positive semidefinite. Namely, the smoothed Instead of filtering each entry of R̃kzz , we now smooth
matrix may not be a valid covariance matrix that should be that of upper triangular matrix Uk across tones by the
positive semidefinite containing Mr2 constrains1 . filtering matrix P. After the smoothing, we reconstruct the
spatial covariance for each tone as R̂kzz = (Ûk )H Ûk . This
reconstruction guarantees that R̂kzz is (Hermitian and) positive
B. Cholesky Decomposition
Ûk e = f H f ≥ 0
semidefinite because eH R̂kzz e = (Ûk e)H 
In multivariate statistics, it is a common approach to decom- f
pose the complicated covariance matrices into simpler compo- for any vector e. The correlation in each entry of Uk across
nents for further processing. There are three popular choices tones can be exploited by linear minimum mean square error
for matrix decomposition: variance-correlation decomposition, (MMSE) filter (i.e. Wiener filter). Other choice of smooth
spectral decomposition (singular value decomposition (SVD)) function might be possible, e.g. Kaiser-Bessel window. The
and Cholesky decomposition. While the entries of the cor- algorithm is summarized in Table I. We show the averaged
relation and orthogonal matrices in the variance-corrleation accuracies (across tones) of different spatial covariance esti-
and spectral decompositions are still constrained, those of Uk mation approaches in Fig 1.
in the Cholesky decomposition are always unconstrained [9].
This property of Cholesky decomposition can be exploited in III. I NTERFERENCE AWARE R ECEIVER D ESIGN
1 A complex unconstrained M × M matrix has 2M 2 degrees of freedom.
r r
A. MMSE Receiver for Co-channel Interference Mitigation
r
There exists a Cholesky matrix that is upper triangular with Mr positive and We mitigate the interference on the k th tone using MMSE
Mr (Mr − 1)/2 complex unconstrained entries for each positive semidefinite
matrix. Therefore, each positive semidefinite matrix has 2Mr2 − Mr − filter computed from the estimated R̂kzz . Denote the MMSE
H
Mr (Mr −1)
2
× 2 = Mr2 constrains. filter as w, w = R̂−1 H
yy R̂xy = (ĤĤ + R̂zz )−1 Ĥ , where

Authorized licensed use limited to: Univ of Texas at Dallas. Downloaded on September 13, 2009 at 15:27 from IEEE Xplore. Restrictions apply.
IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 7, NO. 12, DECEMBER 2008 4851

the tone index k is suppressed for notational simplicity. be written as


The straightforward estimate of Ryy is E{yyH }, where the  
Rzz (tn ) 0
average is over the received signals ys of the k-th tone Rzz = 0 R∗ . (3)
zz (tn +1)
within one packet. However, this method does not provide
accurate estimates especially for high order modulations. Since
the received signal y is scaled by the random transmitted C. Bound of the Mean Square Error (MSE)
signal x and the number of samples ys is limited per packet,
the straightforward averaging results in a large variation in The MSE is defined as E{x − wy2 }, which quantifies
practice. Hence, to derive the MMSE receiver, the intended the filtered signal quality. The actual MSE is denoted by
signal channel matrix H and interference statistics Rzz are MSE, and the estimated MSE is computed as:
estimated separately.
ˆ = (1 + HH R̂−1
MSE zz H)
−1
. (4)
The receiver can apply the MMSE filter w directly. Al-
ternatively, one may separate the filtering into two steps:
ˆm= 1
interference whitening and MMSE filtering on the whitened The post-equalizer SNR can be written as SNR ˆ m,m −
MSE
interference. The whitening filter is Û−1 . Noting Û−1 is 1 . The post-equalizer SNR of each data stream is used
upper triangular helps to lower the whitening complexity. This to compute the soft information of each coded bit. Hence,
implementation minimizes the changes on the legacy receiver the post-equalizer MSE determines the receiver performance.
that only handles white noise. In this section, we characterize the relationship between the
estimated MSEˆ in (4) and the actual output MSE. The
difference between the two is caused by the estimation error in
B. Enhancements for Space-time Block Coded (STBC) System R̂zz . To focus on the impact of spatial covariance estimation,
Space-time coding is a powerful approach that exploits we assume the channel matrix H is known and consider the
the spatial diversity to combat fading in MIMO wireless SIMO case below, where there is no intra-user interferences.
communications systems. We use the Alamouti code as an Generalization to MIMO case is straightforward.
example, and modify the signal model in (1) to incorporate Theorem 1: The MSE of the MMSE receiver with the
the space-time code by stacking the received signal vectors estimated spatial covariance is upper bounded by:
from time tn to tn + 1 as
 
⎛ y1 (tn ) ⎞ ⎛ H H12
⎞ ⎛ z1 ⎞ ˆ + MSE
MSE ≤ MSE ˆ 2 · ΔRzz F · λmax (R̂−1
ˆ − MSE zz ).
11

⎜ .
.. ⎟ ⎜ .. . .
.. ⎟ . (5)
⎜ ⎟ ⎜ ⎟ ⎜ .. ⎟
⎜ yMr (tn ) ⎟ ⎜ HM∗r 1 HMr∗2 ⎟ x1 ⎜ zMr ⎟
⎜ y1∗ (tn +1) ⎟ = ⎜ H12 −H11 ⎟ ( x2 ) + ⎜ ⎟
⎜ zMr +1 ⎟ . (2) Proof: See Appendix A.
⎜ ⎟ ⎜ ⎟ ⎝ . ⎠
⎝ .. ⎠ ⎝ .. .. ⎠ .. From (5), it is seen that the actual MSE depends on the
. . . ˆ ΔRzz F = Rzz − R̂zz F , and the largest

yM (t +1)
r n

HM r2

−HM r1 
z2Mr
 estimated MSE,
    ˆ 2
y H̃
z eigevalue of R̂−1
zz . We can ignore the second-order term MSE
for reasonable high signal to interference ratio (SIR), where
z is the asynchronous co-channel interference plus noise, ˆ << 1. The second term of the bound includes factors
MSE
which is space-time coded. If the intended and interference of F-norm of the covariance estimation error and the the
signals are synchronized in terms of OFDM cyclic structure largest eigenvalue of R̂−1
zz . ΔRzz F s for different estimation
and space-time modulation, we have 2Mr − 2 degrees of schemes are shown in Fig. 1. The bound implies, besides the
freedom for interference suppression. However, for random F-norm of the estimation error, the eigen-structure of R̂zz has
asynchronous interference, the term z is unstructured. Not also a significant impact to the MSE. In other words, a singular
only the degree of freedom is insufficient, but also we need R̂zz has a large error. Sample covariance matrix, i.e. the
to double the dimension of the “spatial-temporal” covariance ML estimator, has the tendency to spread the eigenvalues [9].
estimation. This prevents us from obtaining good CCI suppres- Therefore, it decreases λmin (R̂zz ) (or increase λmax (R̂−1
zz )),
sion even with the improved covariance estimation techniques and in turn increase the MSE of the equalizer.
in the previous section. The reason is that the asynchrony
makes the space-time coded CCI act like a 2Mr ×2Mr spatial
multiplexing interference. Total 2Mr degrees of freedom at D. MAP Receiver for Co-channel Interference Suppression
the receiver are not enough to suppress the interference signal
effectively. Since we have modeled the interference as Gaussian random
To avoid the high complexity equalization such as decision process with zero mean and covariance Rzz , the optimum
feedback, we propose a heuristic solution that block diagonal- MAP bit detector that minimizes bit error probability can
izes the covariance matrix by zero-forcing the cross correlation be derived. The a priori L-value of the coded bits bi , i =
information between two successively received signal vectors 0, 1, . . . , Nt M − 1, is defined as LA (bi ) = ln PP[b[bi i=−1]
=1]
. To
from time tn to tn + 1. This approximation holds if the compute the LA (bi )s, we can either generate the candidate
asynchrony is not severe or the scrambling sequence varies sets Li,+1 and Li,−1 by exhaustive listing for small antenna
across OFDM symbols. It not only reduces the amount of number and lower modulation order, or generate by the list
estimation parameters by half, but also frees the degrees of sphere decoding for large antenna number and higher order
freedom for interference suppression. More precisely, Rzz can modulation. Interested reader is referred to [10]. Using the

Authorized licensed use limited to: Univ of Texas at Dallas. Downloaded on September 13, 2009 at 15:27 from IEEE Xplore. Restrictions apply.
4852 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 7, NO. 12, DECEMBER 2008

1x2 SIMO, 16 QAM, SNR = 20 dB 2x3 MIMO , Alamouti code, SNR = 20 dB


0 0
10 10
Sync, Inter Channel Known.
Cholesky Decom.+ w/ Block Diag.
Sync, Inter−channel known, MMSE
Cholesky Decom. + w/o Block Diag.
Chol. Decomp&smoothing, MMSE
−1
MRC
Cov smoothing w/o Matrix decomp. MMSE 10
w/o smoothing MMSE
−1
10 MRC
PER

PER
−2
10

−2
10
−3
10

−3 −4
10 10
2 4 6 8 10 12 14 0 2 4 6 8 10 12
SIR (dB) SIR

Fig. 2. Packet error rate of different receivers for 1x2 SIMO, 16 QAM, Fig. 3. Packet error rate for space-time coded system, 2x3 MIMO 16 QAM,
MMSE receiver. MMSE receiver.

max-log approximation and plugging the estimated interfer- dB stronger than that of the MRC for the same PER. The
ence statistics, the extrinsic L-value can be approximated as: straightforward smoothing without Cholesky decomposition
1 loses the positive semidefinite property and only provides 1
LE (bi |y) ≈ max {−||Û−1 (y − Hx)||2 + bT[i] LA,[i] } dB improvement over non-smoothing. In contrast, Cholesky
x∈Li,+1 2
smoothing with the property delivers 3 dB gain over non-
−1 2 1 T
− max {−||Û (y − Hx)|| + b[i] LA,[i] }, smoothing. The PER of the MMSE receiver with synchro-
x∈Li,−1 2
nized interferer and known interfering channel is plotted as
(6)
a benchmark. The asynchrony accounts for a gap less than
where b[i] denotes the sub-vector of b omitting its ith element, 2 dB between asynchronous Cholesky smoothing and the
and LA,[i] is the vector of all LA values, also omitting its ith synchronized case. The MRC curve has a slop steeper than
element. The MAP detector iteratively exchanges the extrinsic those of the MMSE receivers but suffers an SIR loss 5 − 8
information with the outer channel decoder to improve the per- dB. The diversity order of MRC is Mr . The MMSE receivers
formance. Since the MAP detector requires a large candidate works as zeor-forcing receiver at high SNR. Its diversity order
list (exponential to Mt ) to generate the likelihood information is Mr − Ns − Ni + 1 as analyzed in [15], where Ns and Ni
for each bit, the complexity of the MAP detector is higher are the numbers of the intended and interfering spatial streams
than the MMSE receiver, whose complexity per data stream respectively. The diversity order decreases as the number of
is linear to Mt . interferers increases. The SIR loss of MRC is due to the
interference power projected on the intended signal direction.
IV. S IMULATIONS AND R EMARKS For typical high-density WLAN, the performance is dominated
by interfererence instead of diversity.
Gray mapping, 802.11n OFDM symbol level interleaver,
We consider an Alamouti coded system with 2 transmit and
LDPC code, and 802.11n channel model D [13] are employed.
3 receive antennas in Fig. 3. The PERs are plotted for the
The interference is asynchronous with an offset uniformly
schemes with/without block diagonalization, where 6 and 12
distributed within one OFDM symbols (1 − 80 time samples).
silent symbols are used for the diagonalized and undiagonal-
A silent window after the intended preamble is introduced for
ized respectively. As expected, MMSE without diagonization
estimating the spatial covariance, where no signal is sent by
encounters the freedom deficiency problem and degrades the
the intended transmitter. The silent duration is 4 − 12 symbols
performance. The proposed scheme with diagonalized Rzz
for fast interference statistics measurement. Required SIRs for
approaches the performance with synchronized interference
PER 1% are compared. We simulated one interferer case in
and known channel (doted curve). Again, MRC has a better
this paper which was typical in high density LAN from the
diversity gain than those of MMSE cancelation but suffers a
testbed measurement. The more interferers, the less structure
SIR loss of 6 dB for the same reason as before.
of the interference statistics due to the averaging effect and
The performance of MAP decoder with/without iteration are
will cause some interference suppression gain loss.
compared with that of the MMSE receiver in Fig. 4, where
Fig. 2 compares receiver schemes for 1 × 2 SIMO. The
6 silent symbols are used. We assume perfect knowledge
channel matrix of the desired signal is estimated under the
of the intended channel to remove the effect of channel
interference2 and 4 silent symbols are employed. The proposed
estimation and highlight that of the interference covariance
Cholesky smoothing outperforms the conventional MRC by 8
estimation. The MAP demodulator and LDPC decoder it-
dB. Namely, the proposed method can tolerate interference 8
eratively exchange the extrinsic information. The iterations
2 The desired signal channel is estimated under the interference by using deliver a marginal gain of 1 dB over that with a single
the method in [14] with one OFDM preamble symbol. iteration. The MAP receiver successively decodes the packet

Authorized licensed use limited to: Univ of Texas at Dallas. Downloaded on September 13, 2009 at 15:27 from IEEE Xplore. Restrictions apply.
IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 7, NO. 12, DECEMBER 2008 4853

−1
and ΔRzz = Rzz − Rˆzz , Rˆzz
2x4 MIMO, Spatial Multiplexing, SNR = 25 dB
0
10 = UH U and ΔRzz =
H
MAP, 4 Iter. T T. Since 
Soft output MAP UHHH UH F = trace(HHH UH UHHH UH U)
Chol.+smoothing,MMSE
−1 Smoothing, W/O Matrx Decomp. ,MMSE 1
10
W/O Smoothing, MMSE = −1 ,
ˆ
MSE

UTH TUH F = trace(UTH TUH UTH TUH )
−1 −1
PER

= Rˆzz ΔRzz F ≤ ΔRzz F · λmax (Rˆzz ) ,


−2
10

(9)

−3 where we have used the Cauchy-Schwarz inequality


10
trace(AB) ≤ AF · BF , and ABF ≤ BF λmax (A).
Therefore,
 
−4
10
−2 0 2 4 6 8 10
ˆ + MSE
MSE ≤ MSE ˆ 2 · ΔRzz F · λmax (R̂−1
ˆ − MSE zz ).
SIR
(10)
Fig. 4. Packet error rate for 2x4 MIMO system, 16QAM, MAP receiver.

within 1 − 2 iterations for most cases. The MAP receiver


provides more diversity gain than MMSE receiver as expected. R EFERENCES
Surprisingly, MMSE with straightforward smoothing performs [1] J. Zhu, B. Metzler, X. Guo, and Y. Liu, “Adaptive CSMA for scal-
even worse than MMSE without smoothing. Since this does able network capacity in high-density WLAN: a hardware prototyping
not happen in Fig. 2, We believe that it is important to maintain approach,” in Proc INFOCOM, pp. 414-419, Apr. 2006.
the semidefinite property of the smoothed covariance matrix [2] J. Winters, “Optimum combining in digital mobile radio with co-channel
interference,” IEEE J. Select. Areas Commun., vol. 52, pp. 528-538, July
especially for large matrices (e.g. 4 antennas). 1984.
[3] S. Catreux, P. F. Diessen, and L. J. Greenstein, “Attainable throught-
V. C ONCLUSION put of an interference-limited multiple-input multiple-output (MIMO)
We devise an efficient method to estimate the spatial covari- cellular system,” IEEE Trans. Commun., vol. 49, pp. 1307-1311, Aug.
2001.
ance of asynchronous MIMO OFDM interference. The method [4] R. S. Blum, “MIMO capacity with interference,” IEEE J. Select. Areas
consists of a Cholesky decomposition step and a smoothing Commun.,, vol. 21, pp. 793-801, June 2003.
operation over the decomposed matrices across OFDM tones. [5] H. Dai, A. F. Molisch, and H. V. Poor, “Downlink capcity of
interference-limited MIMO systems with joint detection,” IEEE Trans.
It improves the performance for SIMO and space-time coded Wireless Commun., vol. 3, pp. 442-452, June 2003.
systems by more than 5 dB. The output MSE of the receiver [6] Y. Li and N. R. Sollenberger, “Adaptive antenna arrays for OFDM
is analyzed. We also designed the MMSE and MAP receiver systems with cochannel interference,” IEEE Trans. Commun., vol 47,
pp. 217-229, Feb. 1999.
based on the proposed estimation. [7] A. Maltsev, R. Maslennikov, and A. Khoryaev, “Comparative analysis
of spatial covariance matrix estimation methods in OFDM communi-
A PPENDIX cation systems,” in Proc. IEEE Symposium on Signal Processing and
Proof: For SIMO case, H is a column vector. Information Technology (ISSPIT), Aug. 2006.
[8] T. A. Thomas and F. W. Vook, “Asynchronous interference suppression
x̂ = wMMSE · y in broadband cyclic-prefix communication,” in Proc WCNC pp. 568-572,
Mar. 2003.
= HH (HHH + R̂zz )−1 H · x + HH (HHH + R̂zz )−1 · z [9] M. Pourahmadi, M. J. Daniels, and T. Park, “Simultaneous modelling of
the Cholesky decomposition of several covariance matrices,” to appear
(a) HH R̂−1
zz H HH R̂−1
zz · z in J. Multivariate Analysis, 2006.
= ·x+ , (7)
1 + HH R̂−1zz H 1 + HH R̂−1zz H [10] B. M. Hochwald and S. T. Brink, “Achieving near-capacity on a
multiple-antena channel,” IEEE Trans. Commun., vol. 51, pp. 389-399,
where (a) has used the matrix inverse lemma. Hence, Mar. 2003.
MSE = E{(x̂ − x)2 } [11] Q. Li, J. Zhu, X. Guo, and C. N. Geoghiades, “Asynchronous co-channel
interference suppression in MIMO-OFDM systems,” in Proc. ICC 2007,
−1 −1
1 HH Rˆzz Rzz Rˆzz H Glasgow, Scotland.
= −1 + −1
[12] Q. Li, “On the multiple-antenna communication: signal detection, error
(1 + HH Rˆzz H)2 (1 + HH Rˆzz H)2 exponent and quality of service,” Ph.D disseration, Texas A&M Univer-
−1 −1 sity, 2007.
1 HH Rˆzz ΔRzz Rˆzz H [13] TGn channel models IEE802.11-03/940r4, May 2004. [Online]. Availi-
= −1 + −1 able: http://www.802wirelessworld.com.
(1 + HH Rˆzz H) (1 + HH Rˆzz H)2 [14] Y. Li, “Simplified channel estimation for OFDM systems with multiple
  transmit antennas,” IEEE Trans. Wireless Commun., vol. 1, pp. 67-75,
ˆ + MSE
= MSE ˆ 2 · trace HH Rˆzz −1 ΔRzz Rˆzz −1 H Jan. 2002.
[15] J. Winters, J. Salz, and R. D. Gitlin, “The impact of antenna diversity
≤ MSE ˆ 2 · UTH TUH F UHHH UH F ,
ˆ + MSE on the capacity of wireless communication systems,” IEEE Trans.
(8) Commun., vol. 42, pp. 1740-1751, Feb./Apr. 1994.

Authorized licensed use limited to: Univ of Texas at Dallas. Downloaded on September 13, 2009 at 15:27 from IEEE Xplore. Restrictions apply.

You might also like