Professional Documents
Culture Documents
7, JULY 2008
From (50) [10] L. C. Godara and A. Cantoni, “Uniqueness and linear independence
of steering vectors in array space,” J. Acoust. Soc. Amer., vol. 70, pp.
[b; b3 ]H VH ()GGH GGH V()
b 467–475, Aug. 1981.
=0
b3 [11] B. Picinbono, “On circularity,” IEEE Trans. Signal Process., vol. 42,
no. 12, pp. 3473–3482, Dec. 1994.
GH V()
b
= 0: (52)
b3
Stochastic Maximum-Likelihood DOA Estimation in the
Presence of Unknown Nonuniform Noise
Lemma 3 actually shows an equivalence between (26) and the esti-
mator fr (). Chiao En Chen, Student Member, IEEE,
() = [ba(); b3 a()3 ]. From Lemma 3 we know that if Q()
Let a
6
Flavio Lorenzelli, Senior Member, IEEE, Ralph. E. Hudson, and
drops rank at = l , then GH a () = 0. Since A is a full-column- Kung Yao, Life Fellow, IEEE
rank matrix that is orthogonal to G, the following statement can be
6
equivalently made:
• there exists a = l , such that a () is a linear combination of
columns of A(:; q ), q = 1; . . . ; LR + 2LC . Abstract—This correspondence investigates the direction-of-arrival
From Assumption 1, we know the probability for the existence of can (DOA) estimation of multiple narrowband sources in the presence of
nonuniform white noise with an arbitrary diagonal covariance matrix.
be reasonably assumed to be zero. While both the deterministic and stochastic Cramér-Rao bound (CRB)
and the deterministic maximum-likelihood (ML) DOA estimator under
Uniqueness of Estimator (35) this model have been derived by Pesavento and Gershman, the stochastic
6
If there exists a = c and aH ()G1 GH
1 a( ) = 0. Then the
ML DOA estimator under the same setting is still not available in the
literature. In this correspondence, a new stochastic ML DOA estimator
following equations hold: is derived. Its implementation is based on an iterative procedure which
a() 0 concentrates the log-likelihood function with respect to the signal and noise
GH = 0; G H = 0: (53) nuisance parameters in a stepwise fashion. A modified inverse iteration
0 a()3 algorithm is also presented for the estimation of the noise parameters.
Simulation results have shown that the proposed algorithm is able to pro-
vide significant performance improvement over the conventional uniform
6
An equivalent statement is made as follows:
• there exists a = c , such that [aT (); 0T ]T , and [0T ; aH ()]T ML estimator in nonuniform noise environments and require only a few
(:; q ), q = 1; . . . ; LR + iterations to converge to the nonuniform stochastic CRB.
is a linear combination of columns of A
2L C . Index Terms—Direction-of-arrival (DOA) estimation , nonuniform noise,
Similarly, the probability for the existence of is considered as zero, sensor array processing, stochastic maximum likelihood (ML) algorithm.
and the uniqueness for (35) can be guaranteed with probability one.
I. INTRODUCTION
REFERENCES
[1] P. Chage, Y. Wang, and J. Saillard, “A Root-MUSIC algorithm for Direction–of-arrival (DOA) estimation has been one of the central
non circular sources,” in Proc. IEEE Int. Conf. Acoust., Speech, Signal problems in radar, sonar, navigation, geophysics, and acoustic tracking.
Processing (ICASSP), Salt Lake City, UT, May 2001, pp. 7–11.
[2] R. O. Schmidt, “A Signal Subspace Approach to Multiple Source Loca- A wide variety of high-resolution narrowband DOA estimators have
tion and Spectral Estimation,” Ph.D. dissertation, Stanford Univ., Stan- been proposed and analyzed in the past few decades [2]–[5]. The max-
ford, CA, May 1981. imum likelihood (ML) estimator, which shows excellent asymptotic
[3] R. Roy, A. Paulraj, and T. Kailath, “ESPRIT-estimation of signal pa- performance, plays an important role among these techniques. Many of
rameters via rotational invariance techniques,” IEEE Trans. Acoust.,
Speech, Signal Process., vol. 37, pp. 984–995, Jul. 1989.
the proposed ML estimators are derived from the uniform white noise
[4] P. Stoica and K. Sharman, “Maximum likelihood methods for direction assumption [5]–[7], in which the noise process of each sensor is as-
estimation,” IEEE Trans. Acoust., Speech, Signal Process., vol. 38, pp. sumed to be spatially uncorrelated white Gaussian with identical un-
1131–1143, Feb. 1990. known variance. It is shown that under this assumption the estimates of
[5] M. Pesavento, A. B. Gershman, and K. M. Wong, “Direction finding in
the nuisance parameters (source waveforms and noise variance) can be
partly calibrated sensor arrays composed of multiple subarrays,” IEEE
Trans. Signal Process., vol. 50, no. 9, pp. 2103–2115, Sep. 2002. expressed as a function of DOAs [8]–[10], and, therefore, the number of
[6] J. Li, B. Halder, P. Stoica, and M. Viberg, “Computationally efficient independent parameters to be estimated is substantially reduced. This
angle estimation for signals with known waveforms,” IEEE Trans.
Signal Process., vol. 43, no. 9, pp. 2154–2163, Sep. 1995.
[7] A. van der Veen and A. Paulraj, “An analytical constant modulus al- Manuscript received May 21, 2007; revised November 26, 2007. This work
gorithm,” IEEE Trans. Signal Process., vol. 44, no. 5, pp. 1136–1155, was supported by in part by the NSF CENS program NSF Grant EF-0410438,
May 1996. by AROD-MURI PSU Contract 50126, and by ST Microelectronics, Inc.The
[8] H. Abeida and J. P. Delmas, “Music-like estimation of direction of associate editor coordinating the review of this correspondence and approving
arrival for noncircular sources,” IEEE Trans. Signal Process., vol. 54, it for publication was Dr. Eran Fishler.
no. 7, pp. 2678–2690, Jul. 2006. The authors are with the Electrical Engineering Department, University
[9] K. C. Tan, S. S. Goh, and E. C. Tan, “A study of the rank-ambiguity of California, Los Angeles, CA 90095 USA (e-mail: chiaoen@ucla.edu;
issues in direction-of-arrival estimation,” IEEE Trans. Signal Process., lorenz@ucla.edu; ralph@ucla.edu; yao@ee.ucla.edu).
vol. 44, no. 4, pp. 880–887, Apr. 1996. Digital Object Identifier 10.1109/TSP.2008.917364
procedure is called concentration which results in a substantial reduc- source with respect to the array centroid, where m = 1; . . . ; M . The
tion of the parameter space required for numerical optimization. y
array output (t), observed at the tth snapshot can then be modeled as
In many applications, the uniform white noise assumption may be
unrealistic. For densely placed sensors, the sensor noise may be corre-
lated and therefore should be modeled as a colored random process. In y(t) = Ax(t) + e(t); t = 1; . . . ; N (1)
order to reduce the number of nuisance noise parameters, various noise
modeling techniques have been applied in the literature [11]–[15]. In where x(t) = [x1 (t); . . . ; xM (t)]T and e(t) are the signal and the
[11], [12], the noise process is assumed to follow the AR(MA) model noise vector observed at the tth snapshot respectively, and
while in [13], [14] the noise covariance matrix is parameterized to
have a linear structure. For non-Gaussian and nonstationary environ-
ments, the generalized autoregressive conditional heteroscedasticity
A = [a(1); . . . ; a(M )] (2)
(GARCH) modeling has been proposed recently [15]. is the steering matrix. It will be assumed that the geometry of the sensor
In some practical applications, the sensors may be sparsely placed
A
array is such that is full rank for all DOAs of interest. As a result,
so that the noise at each sensor is essentially uncorrelated. However,
the noise power of each sensor can still be different due to the nonuni-
A AH will always be positive definite.
In this paper, we investigate the case where the sensor noise of each
formity of sensor noise or the imperfection of array calibration. In this
case, the noise covariance matrix can be modeled as a diagonal matrix channel is a spatially uncorrelated white Gaussian random process, and
but the diagonal elements in general are not identical. It is crucial to it follows that the noise covariance matrix has the following diagonal
note that this scenario can not be obtained as a special case of the Au- structure:
9^ = arg max
randomly distributed array. For simplicity, we assume the sources and
the array lie in the same plane, and denote m as the DOA of the mth 9
9 ):
L(9 (9)
3040 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 56, NO. 7, JULY 2008
Clearly, the search space of solving (9) is huge (of dimension M + It appears that (19) is a rather complicated function of Q and no ana-
+
M 2 P ), and, therefore, we seek to reduce the optimization problem lytical solution seems to be available.
2
since the estimation of is our only interest. To this end, a new ML DOA estimator based on the stepwise concen-
We first maximize (9) with respect to Rx for fixed and Q. Taking 2 tration is proposed. The idea of this technique is to numerically concen-
partial derivatives of L (99)
with respect to Rx i;j , we obtain [ ] trate the log-likelihood function by the following iterative procedure.
@L (99)
[ ]
@ Rx i;j
Iterative procedure of the proposed ML DOA estimator
^ x = A SA 0 ^ [A A]01
y y
P
R 2
A Ry01 [S 0 Ry ]R0
y 1 A = 0:
H H
Q
where Ay = (A A)01 A .
P
For a more detailed derivation of (10)–(11), see [9]. H H
R (12)
increases the log-likelihood function.
Step 3) Use the obtained Q ^ to find an improved 2^
where ^
through 2 = argmax2 L(2 2; Q^ ) where
~ = Q01=2A L(2 2; Q) is defined as in (16). Update R^ x using
A (13)
the latest estimates (2 ^ ; Q^ ) through (12).
S~ = Q01=2 SQ01=2 : (14) Iter = Iter + 1
Substituting Ry of (5) by ARx AH ^ + Q, the ML estimate for 2
Repeat Step 2 and Step 3 until the
algorithm converges.
and Q can be obtained by solving
N
or equivalently
= trf[0R^ y01 + R^ y01(S 0 1Q)R^ y01]1Qg
[R0y 1 0 Ry01SR0y 1] = 0; p = 1; . . . ; P:
p;p (20) +trf[R^ y011QR^ y01]1Qg (22)
IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 56, NO. 7, JULY 2008 3041
The equality sign in (23) holds since the first term in (22) is zero by
construction.
The condition (21) is equivalent to
Putting (24) into a matrix form, then 1q, the vector of the diagonal
elements of 1Q can be solved by the following linear equation:
H1q = u (25)
where
and
Fig. 2. Comparison of the DOA estimation RMSEs and the stochastic CRB in Fig. 4. Comparison of the DOA estimation RMSEs and the stochastic CRB in
the presence of nonuniform sensor noise (uncorrelated sources scenario). the presence of nonuniform sensor noise (correlated sources scenario).
Fig. 3. Comparison of the DOA estimation RMSEs and the stochastic CRB in Fig. 5. Comparison of the DOA estimation RMSEs and the stochastic CRB
the presence of nonuniform sensor noise (correlated sources scenario). versus the number of sensors.
0.1 (same as in [1] but due to a typo appeared as 0.9). The ASNR is set In the fifth scenario, the RMSE of the tested algorithms versus the
to be 7.5 dB and the RMSE performance is plotted with respect to the number of sensors is investigated. We assume the sources to be uncor-
number of snapshots in Fig. 3. related and the ASNR, number of snapshots, and the WNPR are set to
In the fourth scenario, we assume the same correlated sources as the be 20 dB, 100, and 65, respectively. Simulation results are shown in
third simulation scenario. N is fixed to be 100 and the RMSEs of the Fig. 5.
tested algorithms are plotted in Fig. 4 versus the ASNR. In the sixth scenario, we assume uncorrelated sources again and set
In the following two simulation scenarios, we investigate how the the ASNR, number of snapshots, and the number of sensors to be 0
nonuniformity of the sensor noise affects the performance of the tested dB, 100, and 10, respectively. The RMSE performance is plotted with
algorithms. As in [1], we define the worst-noise-power-ratio (WNPR) respect to the WNPR in Fig. 6.
2
max From Figs. 1–6, it can be observed that all the tested nonuni-
WNPR = (31)
min
2 form algorithms provide essential performance improvement over
the conventional (uniform) ML estimator in the nonuniform noise
2
as a measure of nonuniformity of the sensor noise, where max
and environment. As the nonuniformity of sensor noise becomes larger
2
min are the highest and lowest sensor noise variance in the array, (WNPR 1), a more substantial performance improvement of the
respectively. In each Monte Carlo run, we fix the WNPR and ran- proposed nonuniform ML estimator over the uniform ML estimator is
domly choose two sensor locations in the ULA, one with noise vari- observed. The performance improvement also becomes more signifi-
2 2 2
ance min and the other max = WNPRmin . The rest of the sensors cant as the number of degrees of freedom (N times P ) increases.
U
are assigned their noise variances according to a uniform distribution Among the tested nonuniform algorithms, the PD method has the
2 2
(min ; max ). lowest complexity since the parameter space is only of dimension M
IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 56, NO. 7, JULY 2008 3043
REFERENCES
[1] M. Pesavento and A. M. Gershman, “Maximum-likelihood direc-
tion-of-arrival estimation in the presence of unknown nonuniform
noise,” IEEE Trans. Signal Process., vol. 49, no. 7, pp. 1310–1324,
Jul. 2001.
[2] R. Schmidt, “A signal subspace approach to multiple emitter location
and spectral estimation,” Ph.D. dissertation, Stanford Univ., Stanford,
CA, 1981.
[3] M. Wax and T. Kailath, “Optimum localization of multiple sources by
passive arrays,” IEEE Trans. Acoust., Speech Signal Process., vol. 31,
no. 5, pp. 1210–1218, Oct. 1983.
[4] T. Kailath and R. Roy, “Estimation of signal parameters via rotational
invariance techniques,” IEEE Trans. Acoust., Speech Signal Process.,
vol. 37, pp. 984–995, Jul. 1989.
[5] P. Stoica and A. Nehorai, “Music, maximum likelihood and
Cramer-Rao bound,” IEEE Trans. Acoust., Speech Signal Process.,
vol. 37, pp. 720–741, May 1989.
[6] J. F. Bohme, “Estimation of spectral parameters of correlated signals
in wavefields,” Signal Process., vol. 11, no. 4, pp. 329–337, Dec. 1986.
[7] P. Stoica and A. Nehorai, “Performance study of conditional and
unconditional direction-of-arrival estimation,” IEEE Trans. Acoust.,
Fig. 6. Comparison of the DOA estimation RMSEs and the stochastic CRB Speech Signal Process., vol. 38, pp. 1783–1795, Oct. 1990.
versus the WNPR. [8] J. F. Bohme, “Separated estimation of wave parameters and spectral
parameters by maximum likelihood,” presented at the IEEE Int. Conf.
Acoustics, Speech, Signal Process. (ICASSP), Tokyo, Japan, 1986.
[9] A. G. Jaffer, “Maximum likelihood direction finding of stochastic
and no additional computation is required to update the nuisance pa- sources: A separable solution,” presented at the Int. Conf. Acoustics,
Speech, Signal Process. (ICASSP), New York, Apr. 1988.
rameters [26]. However, the RMSE performance of the PD method ap-
[10] P. Stoica and A. Nehorai, “On the concentrated stochastic likelihood
pears to be the worst and approaches that of the uniform ML estimator function in array signal processing,” Circuits, Syst., Signal Process.,
as the ASNR is sufficiently large. This is not surprising since the ob- vol. 14, pp. 669–674, Sep. 1995.
jective function of PD degenerates to the same ML metric of the deter- [11] J. Le Cadre, “Parametric methods for spatial signal processing in
ministic uniform ML estimator when p2 approaches zero. The RMSE the presence of unknown colored noise fields,” IEEE Trans. Acoust.,
Speech Signal Process., vol. 37, no. 7, pp. 965–983, Jul. 1989.
performance and the complexity of the proposed stochastic nonuniform [12] V. Nagesha and S. Kay, “Maximum likelihood estimation for array
ML estimator is comparable to that of the deterministic nonuniform processing in colored noise,” IEEE Trans. Signal Process., vol. 44, no.
ML estimator [1] since both algorithms implement the idea of step- 2, pp. 169–180, Feb. 1996.
wise-concentration and both require only two iterations to converge to [13] J. F. Bohme and D. Kraus, “On least squares methods for direction of
arrival estimation in the presence of unknown noise field,” presented
a solution close to the CRB in the large sample and high ASNR cases.
at the IEEE Int. Conf. Acoustics, Speech, Signal Process. (ICASSP),
However, the deterministic nonuniform ML estimator seems to have a New York, Apr. 1988.
better threshold property as its RMSE is smaller than that of the sto- [14] B. Friedlander and A. F. Weiss, “Direction finding using noise co-
chastic nonuniform ML estimator in the low ASNR regions. The AML variance modeling,” IEEE Trans. Signal Process., vol. 43, no. 7, pp.
algorithm has a RMSE performance comparable to both the determin- 1557–1567, Jul. 1995.
[15] H. Amiri, H. Amindavar, and R. L. Kirlin, “Array signal processing
istic and stochastic ML estimators at the expense of higher complexity
O
(appears to be around ((M 2 + P )3 ) comparing to the complexity
using GARCH noise modeling,” presented at the IEEE Int. Conf.
Acoustics, Speech, Signal Process. (ICASSP), Montreal, QC, Canada,
O 3
(M ) of PD and the nonuniform ML estimators [26]). However, May 2004.
when N is small, the RMSE performance of AML degrades consid- [16] A. M. Zoubir and S. Aouada, “High resolution estimation of directions
of arrival in nonuniform noise,” presented at the IEEE Int. Conf. Acous-
erably, which is probably due to the asymptotic approximations made
tics, Speech, Signal Process. (ICASSP), Montreal, QC, Canada, May
in the derivations. 2004.
[17] A. B. Gershman, A. L. Matveyev, and J. F. Bohme, “Maximum like-
lihood estimation of signal power in sensor array in the presence of
VI. CONCLUSION unknown noise field,” Proc. Inst. Elect. Eng. Radar, Sonar Navig., pp.
In this correspondence, we address the problem of estimating the 218–224, Oct. 1995.
[18] B. Goransson and B. Ottersten, “Direction estimation in partially un-
DOAs of multiple narrowband sources in the presence of unknown known noise fields,” IEEE Trans. Signal Process., vol. 47, no. 9, pp.
nonuniform sensor noise. A new stochastic ML DOA estimator has 2375–2385, Sep. 1999.
been derived. Unlike the conventional uniform ML DOA estimator, no [19] I. Ziskind and M. Wax, “Maximum likelihood localization of multiple
separate solution for the DOA estimation seems to exist in the nonuni- sources by alternating projection,” IEEE Trans. Acoust., Speech, Signal
Process., vol. 36, no. 10, pp. 1553–1560, Oct. 1988.
form case. To this end, we implement our algorithm by using the so [20] K. Sharman, “Maximum likelihood estimation by simulated an-
called stepwise concentration which concentrates all the nuisance pa- nealing,” in Proc. IEEE Int. Conf. Acoustics, Speech, Signal Process.
rameters iteratively. A modified inverse iteration algorithm has also (ICASSP), New York, pp. 2741–2744.
been derived for the estimation of the noise covariance matrices. [21] P. Stoica and A. B. Gershman, “Maximum-likelihood DOA estimation
by data-supported grid search,” IEEE Signal Process. Lett., vol. 6, pp.
The performance of the proposed algorithm is studied through ex-
273–275, Oct. 1999.
tensive computer simulations. In all settings, the proposed algorithm [22] S. Kay and S. Saha, “Mean likelihood frequency estimation,” IEEE
asymptotically converges to the CRB within two iterations and there- Trans. Signal Process., vol. 48, no. 7, pp. 1937–1946, Jul. 2000.
fore the complexity is only a few times higher than the conventional [23] M. Li and Y. Lu, “Genetic algorithm based maximum likelihood DOA
uniform ML estimator. Simulation results also demonstrate the perfor- estimation,” in Proc. Inst. Elect. Eng. Int. Radar Conf., Edinburgh,
U.K., Oct. 2002.
mance improvement in comparison with other nonuniform algorithms [24] P. J. Chung and J. F. Bohme, “DOA estimation using fast EM and
for a variety of scenarios (number of snapshots, SNR, number of sen- SAGE algorithms,” Signal Process., vol. 82, pp. 1753–1762, Nov.
sors, etc.). 2002.
3044 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 56, NO. 7, JULY 2008
[25] J. P. Burg, D. G. Luenberger, and D. L. Wenger, “Estimation of struc- where X is a perturbed version of the data matrix X , and 1
X is the
tured covariance matrices,” Proc. IEEE, vol. 70, Sep. 1982. perturbation. The subspace decomposition of X is
[26] D. Madurasinghe, “A new DOA estimator in nonuniform noise,” IEEE
Signal Process. Lett., pp. 337–339, Apr. 2005.
[27] A. B. Gershman, P. Stoica, M. Pesavento, and E. G. Larsson, “Sto-
chastic Cramer-Rao bound for direction estimation in unknownnoise X = 6
U V
H = U s6sV Hs + U n6nV Hn (2)
fields,” Proc. Inst. Elect. Eng. Radar, Sonar Navig., pp. 2–8, Feb. 2002.
where U s , associated with nonzero singular values, spans the column
space of X , which is also called the signal subspace, and U n , associ-
ated with zero singular values n (66 = ) 0 , spans the orthogonal space
of U s , which is also called the noise subspace. Similarly, the subspace
First-Order Perturbation Analysis of Singular Vectors in decomposition of X is given by
Singular Value Decomposition
H H H
Jun Liu, Student Member, IEEE,
Xiangqian Liu, Senior Member, IEEE, and
X = 6
U V = s6 s
U V s + U n6nV n : (3)
Xiaoli Ma, Member, IEEE
1
Due to the perturbation X , all the quantities in the right-hand side
(RHS) of (3) may differ from those in the RHS of (2). For example, we
= +1
may express U s as U s U s U s , where 1
U s is the perturbation
Abstract—Singular value decomposition (SVD) is an important tech- of the singular vectors that span the signal subspace.
nique in signal processing. In the literature, the perturbation analysis of Perturbation analysis results for subspace decomposition have been
SVD has been well documented in the context of subspace decomposition. well documented in the literature. For example, the statistics of eigen-
The contribution of the signal subspace to the perturbation of the singular vectors of sample covariance matrices is investigated in [1]. The pertur-
vectors that span the signal subspace is often ignored as it is treated as
bation of the invariant subspace is considered in [2] and [3]. Based on
second and higher order terms, and thus the first-order perturbation is
typically given as the column span of the noise subspace. In this corre- 1
the results in [2] and [3], a first-order approximation of U s is given in
spondence, we show that not only the noise subspace, but also the signal [4]. Furthermore, [5] and [6] give a second-order approximation using
subspace, contribute to the first-order perturbation of the singular vectors. similar arguments as in [4]. It is shown in [4]–[6] that the contribution
We further show that the contribution of the signal subspace does not
1
of the noise subspace spanned by U n to U s is of the first order of
impact on the performance analysis of algorithms that rely on the signal
1 X . It is argued that the contribution of the signal subspace spanned
1 1
subspace for parameter estimation, but it affects the analysis of algorithms
that depends on the individual basis vectors. For the latter, we also give by U s to U s is of the second order of X , and thus the contribution
a condition under which the contribution of the signal subspace to the
perturbation of singular vectors may be ignored in the statistical sense.
1
of the signal subspace to U s is ignored when the first-order pertur-
bation is considered.
Numerical examples are provided to validate our theoretic claims.
In this correspondence, we analyze the perturbation of U s from an-
Index Terms—Perturbation analysis, singular value decomposition other perspective. Different from the existing results in [4]–[6], we find
(SVD), subspace decomposition. 1
that the contribution of the signal subspace to U s is also of the first
1
order of X . The difference is explained in detail in Remarks 1 and
2 of Section II. Furthermore, we show that the signal subspace has an
I. INTRODUCTION impact on the first-order perturbation of the individual singular vectors,
but not on the first-order perturbation of the signal subspace spanned
Singular value decomposition (SVD) plays an important role in by these vectors. For this reason, this new result has no material effect
signal processing. For example, subspace-based algorithms in array on the performance analysis of subspace-based algorithms that utilize
signal processing depends on SVD or eigenvalue decomposition the signal subspace, e.g., ESPRIT [7], MUSIC [8], the column span
to obtain a basis of the desired signal and/or noise subspaces, and methods [9], as well as the subspace-based algorithms discussed in
principal component analysis (PCA) uses SVD to retrieve principal [4]–[6]. In other words, the same results will be obtained in the per-
components from noisy observations. Therefore, perturbation analysis formance analysis of these algorithms no matter whether the contri-
of SVD is important for these algorithms. Consider the following
observation matrix
bution of the signal subspace to the first-order approximation of U s 1
is considered or not. However, our result does affect the performance
analysis of methods that depend on the individual singular vectors,
X = +1
X X (1) e.g., prewhitening in JADE [11] and the principal component analysis
(PCA) [10].
The rest of the correspondence is organized as follows. The perturba-
Manuscript received February 20, 2007; revised November 29, 2007. The tion of the singular vectors of a perturbed matrix is given in Section II.
associate editor coordinating the review of the manuscript and approving it for In Section III, we give a condition under which the contribution of the
publication was Dr. Sergiy A. Vorobyov. This work was supported in part by
the U.S. Army Research Laboratory and the U.S. Army Research Office under
1
signal subspace to the first-order approximation of U s can be ignored
grant number W911NF-06-1-0415 and W911NF-06-1-0090. Parts of this work in the statistical sense. In Section IV, we apply the perturbation result
were presented at the IEEE Statistical Signal Processing Workshop, Madison, to analyze several subspace-based algorithms. Numerical examples are
WI, August 26–29, 2007. provided in Section V to validate the theoretic claims.
J. Liu and X. Liu are with Department of Electrical and Computer En- In the following, upper (lower) boldface letters are used for matrices
gineering, University of Louisville, Louisville, KY 40292 USA (e-mail: (column vectors). A 3 , A T , and AH denote the conjugate, transpose,
jun.liu@louisville.edu; x.liu@louisville.edu).
and Hermitian transpose of A , respectively. We will use I M for an
identity matrix of size M 2 M , 0M 2N for a zero matrix of size M 2 N ,
X. Ma is with School of Electrical and Computer Engineering, Georgia Insti-
tute of Technology, Atlanta, GA 30332 USA (e-mail: xiaoli@ece.gatech.edu).
Digital Object Identifier 10.1109/TSP.2007.916137 kAk for the Frobenius norm of A, and kAk2 for the 2 norm of A.
1053-587X/$25.00 © 2008 IEEE