 Research
 Open Access
 Published:
Multichannel adaptive signal detection in spacetime colored compoundgaussian autoregressive processes
EURASIP Journal on Advances in Signal Processing volume 2012, Article number: 69 (2012)
Abstract
In this article, we consider the problem of adaptive detection for a multichannel signal in the presence of spatially and temporally colored compoundGaussian disturbance. By modeling the disturbance as a multichannel autoregressive (AR) process, we first derive a parametric generalized likelihood ratio test against compoundGaussian disturbance (CGPGLRT) assuming that the true multichannel AR parameters are perfectly known. For the twostep GLRT design criterion, we combine the multichannel AR parameter estimation algorithm with three covariance matrix estimation strategies for compoundGaussian environment, then obtain three adaptive CGPGLRT detectors by replacing the ideal multichannel AR parameters with their estimates. Owing to treating the random texture components of disturbance as deterministic unknown parameters, all of the proposed detectors require no a priori knowledge about the disturbance statistics. The performance assessments are conducted by means of Monte Carlo trials. We focus on the issues of constant false alarm rate (CFAR) behavior, detection and false alarm probabilities. Numerical results show that the proposed adaptive CGPGLRT detectors have dramatically ease the training and computational burden compared to the generalized likelihood ratio testlinear quadratic (GLRTLQ) which is referred to as covariance matrix based detector and relies more heavily on training.
1 Introduction
In an airborne radar system, spacetime adaptive processing (STAP) has been widely used in radar target detection; see [1–4] and references therein. Various wellknown STAP based detectors have been extensively investigated under Gaussian assumption [4–7]. However, with the support of measured data, the Gaussian model is no longer suitable for background disturbance in many situations of practical interest. The conventional STAP detectors may suffer severe performance degradation when the disturbance is nonGaussian. Instead, a compoundGaussian model can successfully describe the nonGaussian disturbance as a product of a spatially and temporally "slowly varying" texture and a locally "rapidly varying" Gaussian speckle component [8, 9]. The texture component accounts for random power variations over range cells. This model includes the socalled spherically invariant random processes (SIRPs). Working with the compoundGaussian model, a multitude of adaptive detectors have been studied in the past few years, for instance, the generalized likelihood ratio testlinear quadratic (GLRTLQ) which was independently derived in [10, 11], the detectors with Rao and Wald tests [12], the Bayesian optimum radar detector (BORD) [13], and so forth. Notice that all of the aforementioned STAP detectors proposed in Gaussian and compoundGaussian environment can be considered as covariance matrix based detectors [8, 14]. Implementing these detectors involves estimating and inverting a spacetime covariance matrix of the disturbance signal for each cell under test (CUT) utilizing independent and identically distributed (IID) targetfree training data (or secondary data). Obviously, when the joint spatialtemporal dimension is large, the training and computational requirements will be quite onerous. Moreover, some practical situations may exacerbate the training data selection and collection problem and limit the amount of appropriate IID training data. The lack of training may lead to illconditioned covariance matrix estimate and significant degradation in the covariance matrix based detection procedure.
To overcome the above difficulty caused by large joint spatialtemporal dimension, the structural information about the disturbance spacetime covariance matrix can be exploited. More precisely, a multichannel autoregressive (AR) process has been found to be able to model the spatialtemporal correlation of the disturbance efficiently [15–18]. In [15], based on approximating the disturbance spectrum with a multichannel AR model of low order, a parametric adaptive matched filter (PAMF) for STAP detection was presented for multichannel system in Gaussian environment. The PAMF detector, which has been proved to be equivalent to a parametric Rao detector in [16], has dramatically outperformed the conventional adaptive matched filter (AMF) [6, 7] with small training size. Also in Gaussian background, a parametric GLRT [17] and a simplified parametric GLRT [18] have been successively developed by utilizing a parametric model (multichannel AR model) in the GLR principle. Experimental results on simulated and real data show that two parametric GLRT detectors work well with limited or even no range training data [18–20]. However, under such conditions, the AMF detector and Kelly's GLRT cannot be implemented. Moreover, application of the multichannel AR model in nonstationary Gaussian clutter for STAP are investigated in [21–23].
For the corresponding problem in compoundGaussian environment, a nonGaussian parametric adaptive matched filter (NGPAMF) has been derived in [24]. However, this test involves explicit knowledge of the disturbance statistics, which are not always available. Unlike the NGPAMF, the normalized parametric adaptive matched filter (NPAMF) reported in [8, 14] requires no a priori knowledge about the disturbance statistics. This feature is rather important in realtime operation. However, Michels et al. [14] still combined the multichannel AR identification algorithm with the sample covariance matrix (SCM) even in compoundGaussian background. The sample matrix is the maximum likelihood estimate (MLE) of the covariance matrix for Gaussian disturbance, but is no longer the MLE for compoundGaussian disturbance. The covariance matrix estimation in compoundGaussian environment is generally intractable. Conte et al. have advocated the use of a normalized sample covariance matrix (NSCM) in [25]. Also, considering the texture component as an unknown deterministic quantity, an approximate ML (AML) estimator has been derived by Gini et al. [26].
Motivated by the previous studies, the main purpose of this article is to derive a parametric GLRT (PGLRT) for detecting a multichannel signal in the presence of compoundGaussian disturbance modeled as a multichannel AR process. Without any knowledge about the disturbance statistics, we resort to a suboptimal GLRT algorithm considering the texture components as unknown deterministic parameters. In further derivation, to get round the difficulty in performing the joint maximization for all the unknown parameters, a twostep GLRT design criterion is adopted in this article. We first derive the modelbased parametric GLRT in compoundGaussian environment (CGPGLRT), which possesses the perfect knowledge about the multichannel AR parameters. We apply three covariance matrix estimation strategies, i.e., sample covariance matrix (SCM), normalized sample covariance matrix (NSCM) and approximate ML (AML) estimator, to the multichannel AR parameter estimation procedure for estimating the unknown AR coefficient matrices. Then three adaptive versions of the CGPGLRT detector: CGPGLRTSCM, CGPGLRTNSCM, and CGPGLRTAML are obtained. Finally, the performance assessments are presented. Numerical results indicate that the CGPGLRTSCM detector has no texture CFAR property, while all of the CGPGLRT, CGPGLRTNSCM, and CGPGLRTAML detectors ensure CFAR property with respect to the texture probability density function (PDF). Compared to the covariance matrix based detector, all of the proposed adaptive detectors can handle the traininglimited case and significantly decreased the computation complexity in compoundGaussian environment.
This article is organized as follows. The problem statement and the signal and disturbance models are presented in Section 2. The parametric GLRT detector for compoundGaussian environment (CGPGLRT) and adaptive CGPGLRT detectors are derived in Section 3. Then the performance assessment of the proposed detectors is displayed in Section 4. Finally, conclusions are given in Section 5.
Notation: Vectors (matrices) are denoted by boldface lower (upper) case letters, all vectors are column ones, superscripts (•) ^{T} and (•) ^{H} denote transpose and complex conjugate transpose, respectively, $\mathcal{C}\mathcal{N}\text{(}\mu \text{,}R\text{)}$ denotes the multivariate complex Gaussian distribution with mean μ and covariance matrix R. denotes the complex number field and det{•} takes the determinant of a matrix.
2 Problem statement and signal model
2.1 Problem statement
Consider the scene that the radar transmits a coherent train of N pulses and receives the signal with a uniform linear array with J sensors. The received data collected over K range cells is organized in a J × N × K data cube. For the range cell under test (CUT), a binary hypothesis test is applied to the JNdimensional complex baseband spacetime vector of primary data ${x}_{0}={\left[{x}_{0}^{T}\left(0\right),{x}_{0}^{T}\left(1\right),\dots ,{x}_{0}^{T}\left(N1\right)\right]}^{T}\in {\u2102}^{JN\times 1}$. Typically, x_{0} contains an unwanted additive disturbance signal ${d}_{0}={\left[{d}_{0}^{T}\left(0\right),{d}_{0}^{T}\left(1\right),\dots ,{d}_{0}^{T}\left(N1\right)\right]}^{T}\in {\u2102}^{JN\times 1}$ with unknown spacetime covariance matrix $R\in {\u2102}^{JN\times JN}$ and may contain a target signal α s with deterministic but unknown complex amplitude, α, and known target spacetime steering vector $s={\left[{s}^{T}\left(0\right),{s}^{T}\left(1\right),\dots ,{s}^{T}\left(N1\right)\right]}^{T}\in {\u2102}^{JN\times 1}$. The spacetime steering vector takes the form of the Kronecker product of the normalized spatial and temporal steering vectors. K IID complex baseband spacetime vectors of targetfree training data ${\left\{{x}_{k}={d}_{k}\right\}}_{k=1}^{K}\in {\u2102}^{JN\times 1}$ exist for assisting the signal detection.
The detection problem at hand can be expressed as the following binary hypotheses test:
where
The disturbance signals ${\left\{{d}_{k}\right\}}_{k=0}^{K}$ lump clutter, jamming, and thermal noise, and may be correlated in space and time.
2.2 Signal model
Herein, a multichannel AR process is applied to model the disturbance signal d _{ k }(n) at time n, k = 0, 1, . . . , K:
where ${\left\{{\text{A}}^{H}\left(p\right)\right\}}_{p=1}^{P}$ denotes the unknown J×J coefficient matrices for AR process d _{ k }(n) of known order P, and ε_{ k } (n) denotes the J × 1 spatial noise driving vectors that are temporally white but spatially colored.
The nonGaussian driving process ε_{ k } (n) is modeled as a compoundGaussian process. It follows that ε_{ k } (n) can be thought of as zeromean spherically invariant random vectors (SIRVs), i.e., they can be written in the form ${\epsilon}_{k}\left(n\right)=\sqrt{{\tau}_{k}}\cdot {z}_{k}\left(n\right),k=0,1,\dots ,K,n=0,1,\dots ,N1$. Here the speckle z _{ k }(n) are J × 1 complex, zeromean, Gaussian vectors with unknown covariance matrix Q. The texture component τ_{ k } is a positive random variable over range, but constant over time when it has long temporal coherent. The texture PDF f_{ τ } (τ) is defined to be the characteristic PDF of the complex SIRV. Given a specific value of τ_{ k } , we have ${\epsilon}_{k}\left(n\right){\tau}_{k}~\mathcal{C}\mathcal{N}\left(0,{\tau}_{k}\text{Q}\right)$. The covariance matrix of the driving process is E{ε_{ k } (n)ε_{ k } (n) ^{H} } = E(τ_{ k } )Q where the mean value E(τ_{ k } ) is also the average disturbance power.
Utilizing the multichannel parametric model allows signal whitening adopting a vector onestep linear prediction error filter (PEF), involving temporal whitening via an inverse movingaverage (MA) filter followed by spatial whitening [27]. Thus, the temporally whitened versions of the steering vector, the primary data and the secondary data are obtained as following:
It is worth noting that the P thorder linear prediction coefficients ${\left\{{\text{A}}^{H}\left(p\right)\right\}}_{p=1}^{P}$ are identically equal to the AR(P) process coefficients.
The P thorder predictor error vector is given by
and
respectively, where α = 0 under H_{0} and α ≠ 0 under H_{1}. Equations (7) and (8) imply that ${\stackrel{\u0303}{\text{x}}}_{k}(n)~\text{SIRV[0,}\text{Q},{f}_{\tau}({\tau}_{k})]$ and ${\stackrel{\u0303}{\text{x}}}_{0}\left(n\right)~\text{SIRV[}\alpha \stackrel{\u0303}{\text{s}}\left(n\right),\text{Q},{f}_{\tau}\left({\tau}_{0}\right)]$, where SIRV[μ,Q,f_{ τ } (τ)] denotes a complex SIRV obtained by sampling a SIRP with mean μ, covariance matrix Q, and a characteristic PDF f_{ τ } (τ). As the distribution f_{ τ } (τ) is unknown, the temporally whitened vectors ${\stackrel{\u0303}{\text{x}}}_{k}\left(n\right),k=0,1,\dots ,K$ can be modeled as conditionally Gaussian with the unknown variance τ_{ k } , i.e., ${\stackrel{\u0303}{\text{x}}}_{k}\left(n\right){\tau}_{k}~\mathcal{C}\mathcal{N}\left(0,{\tau}_{k}\text{Q}\right)$ and ${\stackrel{\u0303}{\text{x}}}_{0}\left(n\right){\tau}_{0}~\mathcal{C}\mathcal{N}\left(\alpha \stackrel{\u0303}{\text{s}}\left(n\right),{\tau}_{0}\text{Q}\right)$.
For large data records, the first P values ${\left\{{\text{x}}_{k}\left(n\right)\right\}}_{n=0}^{P1},k=0,1,\dots ,K$ can be ignored [28]. Thus, the joint conditional PDF or likelihood function fi(x _{ k }(0),x _{ k }(1), . . . , x _{ k }(N  1)τ_{ k } ;α,A^{H} ,Q) under hypothesis H_{ i }, i = 0 or 1, can be suitably approximated as
where ε_{ k } (n), for k = 1, . . . , K, is a temporally whitened form of the training signals given by (6) and (7), whereas ε_{0}(n) is given by (4), (5), and (8) with α = 0 under H_{0} and α ≠ 0 under H_{1}. Notice that the secondary data ${\left\{{\text{x}}_{k}\right\}}_{k=1}^{K}$ and the primary data x_{0} are independent. Define $\text{X}\left(n\right)={\left[{\text{x}}_{0}^{T}\left(n\right),{\text{x}}_{1}^{T}\left(n\right),\dots ,{\text{x}}_{K}^{T}\left(n\right)\right]}^{T}$. The final joint conditional PDF is given by
where τ = [τ_{0}, τ_{1}, . . . , τ_{ K } ] ^{T} and ${\text{A}}^{H}=\left[{\text{A}}^{H}\left(1\right),{\text{A}}^{H}\left(2\right),\dots ,{\text{A}}^{H}\left(P\right)\right]\in {\u2102}^{J\times JP}$. Note that A^{H} (p) is expressed in terms of the Hermitian operation for notational convenience, but is not necessarily a Hermitian matrix.
3 Parametric GLRT derivation in compoundgaussian environment
In the NeymanPearson sense, the optimum solution for the composite hypothesis testing problem (1) is the likelihood ratio test (LRT). But for the case at hand, it cannot be implemented due to total ignorance of the signal parameter α, the multichannel AR parameters Q and A^{H} and the texture PDF f_{ τ } (τ). Hence, we resort to a suboptimal GLRT algorithm where the τ_{ k } s are modeled as unknown deterministic parameters, and perform the maximum likelihood estimation (MLE) for all the unknown parameters under each hypothesis. Unfortunately, the exact maximization with respect to the unknown parameters is rather difficult and does not exist a closeform expression. Therefore, to get round the above difficulty, the twostep GLRT design criterion is adopted. We first assume that the AR parameters Q and A^{H} are perfectly known, and derive the CGPGLRT detector based on the primary data. The adaptive versions of CGPGLRT detector are then obtained by substituting the unknown AR parameters with their estimates based on the training signals only.
3.1 CGPGLRT detector for known AR parameters
The GLRT for the case at hand, under the assumption that Q and A^{H} are known as well as τ_{ k } s are modeled as unknown deterministic parameters, is given by
where η is the threshold value to be set in order to ensure the desired probability of false alarm (P_{fa}). Under H_{1} and H_{0}, the joint conditional PDFs of the primary data are
and
respectively.
The ML estimates of τ_{0} are easily derived, from the definition
i = 0 or 1,
where
and tr{·} denotes the trace operator.
Utilizing the two estimators in (14) and (15), the joint conditional PDFs under H_{1} and H_{0}, respectively, are:
Performing the maximization of (18) with respect to α is tantamount to minimizing the following expression:
When the positive factor containing α is made to vanish, the minimum is clearly attained
with
Substituting $\widehat{\alpha}$ as well as ${\widehat{\tau}}_{{0}_{0}}$ and ${\widehat{\tau}}_{{0}_{1}}$ into the decision rule (11), after some algebra manipulations, the detection statistic for parametric GLRT in compoundGaussian environment (CGPGLRT) can be recast as
where η_{CGPGLRT} is the appropriate modification of the original threshold in (11).
Note that the normalized parametric adaptive matched filter (NPAMF), originally developed in [8, 14] for compoundGaussian environment, is closely related to the CGPGLRT detector but replaces true Q and A^{H} with their estimates obtained from the training signals. In [8, 14, 29], several multichannel parameter estimation algorithms are considered in the NPAMF detector, including the NuttallStrand and the multichannel least squares methods [30]. However, we obtain herein adaptive version of the CGPGLRT detector relying on maximum likelihood (ML) parameter estimation criterion. Specifically, our multichannel parameter estimation approach is similar to that proposed in [16, 17] for Gaussian parametric modelbased STAP detectors, whereas we utilize only training signals for parameter estimation.
3.2 Adaptive CGPGLRT detector for unknown AR parameters
To make the derived detector (22) fully adaptive, suitable estimates of the multichannel AR parameters Q and A^{H} must be acquired. Employing the targetfree training signals to estimate the multichannel AR parameters allows to decouple from estimation of target parameters, such as complex amplitude α and the texture component τ_{0}. However, derivation of the MLEs of Q and A^{H} with K unknown textures ${\left\{{\tau}_{k}\right\}}_{k=1}^{K}$ is still a more challenging task. The similar covariance estimation problem in compoundGaussian environment has been solved in [25, 26, 31], where three different estimation strategies are introduced, including SCM, NSCM, and AML estimators.
In our opinion, the derivation of these three covariance estimators depends upon two different clutter models, that is, the dependent interference model[31] and the independent interference model[26], respectively. The former assumes that the textures of training signals are completely correlated, i.e., τ_{1} = τ_{2} = ··· = τ_{ K }, and adopts the SCM estimator. The latter, where the NSCM and AML estimators are usually applied, deals with the IID ${\left\{{\tau}_{k}\right\}}_{k=1}^{K}$ instead.
For our problem, as suggested in [26], a twostep personbyperson maximization is followed. Specifically, we first assume that the textures of the training signals are perfectly known for the realization under observation. Define the training data set ${\text{X}}_{t}\left(n\right)={\left[{\text{x}}_{1}^{T}\left(n\right),{\text{x}}_{2}^{T}\left(n\right),\dots ,{\text{x}}_{K}^{T}\left(n\right)\right]}^{T},n=0,\phantom{\rule{0.3em}{0ex}}1,\dots ,N1$. Conditioned on the first P values ${\left\{{\text{x}}_{k}\left(n\right)\right\}}_{n=0}^{P1},k=1,2,\dots ,K$, the loglikelihood function based on training signals is proportional to
where ${\epsilon}_{k}\left(n\right)={\stackrel{\u0303}{\text{x}}}_{k}\left(n\right)$ and ${\text{D}}_{{\tau}_{t}}\in {\u2102}^{K\times K}$ is a diagonal matrix whose diagonal entries are τ_{1}, . . . , T_{ K } . We, next, use (23) to derive the MLEs of Q and A^{H} .
By setting to zero the derivative of (23) with respect to Q, we get the MLE of Q given τ_{1} = [τ_{1}, . . . , τ_{ K } ] ^{T} and A^{H} :
where ${y}_{k}(n)=\left[{x}_{k}^{T}(n1),\dots ,{x}_{k}^{T}{(nP)}^{T}\right]\in {\u2102}^{JP\times 1}$ and L = (N  P) × K. Substituting the above $\widehat{\text{Q}}\left({\tau}_{t},{\text{A}}^{H}\right)$ back in (23), we find that maximizing (23) with respect to A^{H} reduces to minimizing $\text{det}\left\{\widehat{\text{Q}}\left({\tau}_{t},{\text{A}}^{H}\right)\right\}$. Therefore, the MLE of the AR coefficients matrix A^{H} can be obtained by minimizing $\text{det}\left\{\widehat{\text{Q}}\left({\tau}_{t},{\text{A}}^{H}\right)\right\}$ with respect to A^{H} .
We further observe that [17]
where the above correlation matrices conditioned on τ_{ t } = [τ_{1}, . . . , τ_{ K }] ^{T} are expressed as
and ${\widehat{\text{R}}}_{xy}\left({\tau}_{t}\right)={\widehat{\text{R}}}_{yx}^{H}\left({\tau}_{t}\right)$. Since ${\widehat{\text{R}}}_{yy}\left({\tau}_{t}\right)$ is nonnegative definite and the second and third terms in (25) do not depend on A^{H} , it follows that [17]
where
Inserting (30) into (25), we finally have
Obviously, once the MLE of τ_{ t } = [τ_{1}, . . . , τ_{ K }] ^{T} , namely ${\widehat{\tau}}_{t\text{ML}}={\left[{\widehat{\tau}}_{\text{1ML}},\dots ,{\widehat{\tau}}_{K\text{ML}}\right]}^{T}$, is available, the exact MLEs of A^{H} and Q can be readily obtained according to (30) and (31). However, so far the explicit solutions for ${\widehat{\tau}}_{t\text{ML}}$ are still inaccessible. Therefore, in the second step of the personbyperson maximization, we introduce the aforementioned two clutter models to motivate the use of the three different covariance estimation strategies.
For the dependent interference model, it is interesting that the MLE of Q with identical textures is proportional to the sample covariance matrix estimate of the temporally whitened training signals within a scalar multiplicative factor λ[26, 31]. More precisely, based on the classical sample covariance matrix (SCM) estimator ${\widehat{\text{Q}}}_{\text{SCM}}\left({\text{A}}^{H}\right)=\left(1/L\right){\sum}_{k=1}^{K}{\sum}_{n=P}^{N1}{\stackrel{\u0303}{\text{x}}}_{k}\left(n\right){\stackrel{\u0303}{\text{x}}}_{k}^{H}\left(n\right)$, ${\widehat{\text{Q}}}_{\text{ML}}$ for the dependent interference model is given as
Likewise, following the derivation of (25), (30), and (31), the MLEs of A^{H} and Q for the dependent model can be straightforwardly expressed as
where
Notice that an explicit solution for λ is not necessary since the AR coefficients estimate ${\widehat{\text{A}}}_{\text{SCM}}^{H}$ is λindependent and the adaptive version of detector (22) only requires knowledge of ${\widehat{\text{Q}}}_{\text{SCM}}\left(\lambda \right)$ (λ) within a scale multiplicative constant.
Define ${\widehat{\stackrel{\u0303}{\text{x}}}}_{k}\left(n\right)$ and $\widehat{\stackrel{\u0303}{\text{s}}}\left(n\right)$ are the temporally whitened versions of x _{ k }(n) and s(n), respectively, using ${\widehat{\text{A}}}_{\text{SCM}}^{H}$
Thus, we have $\widehat{\stackrel{\u0303}{\text{S}}}=\left[\widehat{\stackrel{\u0303}{\text{s}}}\left(P\right),\widehat{\stackrel{\u0303}{\text{s}}}\left(P+1\right),\dots ,\widehat{\stackrel{\u0303}{\text{s}}}\left(N1\right)\right]\in {\u2102}^{J\times \left(NP\right)}$ and ${\widehat{\stackrel{\u0303}{\text{X}}}}_{k}=\left[{\widehat{\stackrel{\u0303}{\text{x}}}}_{k}\left(P\right),{\widehat{\stackrel{\u0303}{\text{x}}}}_{k}\left(P+1\right),\dots ,{\widehat{\stackrel{\u0303}{\text{x}}}}_{k}\left(N1\right)\right]\in {\u2102}^{J\times \left(NP\right)}$.
Plugging ${\widehat{\text{Q}}}_{\text{SCM}},\phantom{\rule{2.77695pt}{0ex}}\phantom{\rule{2.77695pt}{0ex}}\widehat{\stackrel{\u0303}{\text{S}}}$, and ${\widehat{\stackrel{\u0303}{\text{X}}}}_{0}$ in place of Q, $\stackrel{\u0303}{\text{S}}$, and ${\stackrel{\u0303}{\text{X}}}_{0}$ into (22), respectively, leads to our adaptive CGPGLRTSCM detector given by
Actually, the dependent interference model can be considered as a simple extension of the partiallyhomogeneous environment under the assumption of the deterministic texture component. Recently, two different parametric Rao tests, referred to as the normalized parametric Rao (NPRao) test and the scaleinvariant parametric Rao (SIPRao) test, respectively, have been developed in [32] for the partiallyhomogeneous environment. The above parametric Rao tests can be candidates for multichannel parametric detection in the dependent interference model. Different to the CGPGLRTSCM detector, the NPRao and SIPRao tests use both training and test signals for parameter estimation. However, the above techniques are no longer applicable to the independent interference model because of the model mismatch.
In the case of the independent interference model, the texture PDF f_{τ}(τ) is not always known a prior. To cope with this uncertainty, the random clutter texture component is treated as a deterministic and unknown parameter. The first candidate for the independent model is the normalized sample covariance matrix (NSCM) estimator [26, 33]. Here, we normalize each term $\left\{{\stackrel{\u0303}{\text{x}}}_{k}\left(n\right){\stackrel{\u0303}{\text{x}}}_{k}^{H}\left(n\right)\right\}$ in the sums in (24) by the datadependent normalization factor ${\widehat{\tau}}_{{k}_{\text{NSCM}}}$ in place of true τ _{ k } to obtain
where ${\widehat{\tau}}_{{k}_{\text{NSCM}}}=\frac{\text{tr{}{\tilde{\text{X}}}_{k}{\tilde{\text{X}}}_{k}^{H}\}}{J(NP)}$ is the sample estimate of the local disturbance power in the reference range cells.
According to (30) and (31), we can also obtain ${\widehat{\text{A}}}_{\text{NSCM}}^{H}$ and ${\widehat{\text{Q}}}_{\text{NSCM}}$ as
where the covariance matrices are also normalized by ${\widehat{\tau}}_{{k}_{\text{NSCM}}}$
From (41) and (42), it is worth observing that an interdependent relationship exists between the estimation of ${\widehat{\tau}}_{{k}_{\text{NSCM}}}$ and ${\widehat{\text{A}}}_{\text{NSCM}}^{H}$. Therefore, it is rather difficult to perform joint estimation for them. The idea pursued here is to obtain ${\widehat{\tau}}_{{k}_{\text{NSCM}}}$ by utilizing ${\widehat{\text{A}}}_{\text{SCM}}^{H}$, namely, we get
In like manner, the adaptive CGPGLRTNSCM detector is given by
where the matrices $\stackrel{\u2323}{\stackrel{\u0303}{\mathbf{\text{S}}}}\in {\u2102}^{J\times \left(NP\right)}$ and ${\stackrel{\u2323}{\stackrel{\u0303}{\mathbf{\text{X}}}}}_{0}\in {\u2102}^{J\times \left(NP\right)}$ consist of the temporally whitened versions of ${\left\{\text{x}\left(n\right)\right\}}_{n=P}^{N1}$ and ${\left\{\text{s}\left(n\right)\right\}}_{n=P}^{N1}$, respectively, using the coefficient estimate ${\widehat{\text{A}}}_{\text{NSCM}}^{H}$.
The second candidate for the independent model is the constrained approximate ML (AML) estimator, provided by the ML theory. Summarily, the constrained AML estimates of Q and A^{H} are derived by jointly solving iteratively the following equations:
for i = 0, 1, 2, . . . , N_{it}, where N_{it} is the number of iterations. Equation (52) guarantees the constraint $\text{tr{}{\widehat{\text{Q}}}_{\text{AML}}(i+1)\}=\text{tr{}\text{Q}\}$ to be satisfied at each iteration. Note that the covariance matrices ${\stackrel{\u0304}{\text{R}}}_{xx},{\stackrel{\u0304}{\text{R}}}_{yy}$, and ${\stackrel{\u0304}{\text{R}}}_{yx}$ have the same structure as those in (44)(46), the only difference being that the normalization factor ${\widehat{\tau}}_{{k}_{\text{NSCM}}}$ is replaced by ${\widehat{\tau}}_{{k}_{\text{AML}}}\left(i+1\right)$ which is refreshed at each iteration. Here, ${\widehat{\text{Q}}}_{\text{NSCM}}$ and ${\widehat{\text{A}}}_{\text{NSCM}}^{H}$ are used as the initialization matrix for this recursive estimator, i.e., ${\widehat{\text{Q}}}_{\text{AML}}\left(0\right)={\widehat{\text{Q}}}_{\text{NSCM}}$ and ${\widehat{\text{A}}}_{\text{AML}}^{H}\left(0\right)={\widehat{\text{A}}}_{\text{NSCM}}^{H}$.
The final estimates ${\widehat{\text{Q}}}_{\text{AML}}\left({N}_{\text{it}}\right)$ and ${\widehat{\text{A}}}_{\text{AML}}^{H}\left({N}_{\text{it}}\right)$ are denoted by ${\widehat{\text{Q}}}_{\text{AML}}$ and ${\widehat{\text{A}}}_{\text{AML}}^{H}$ for the sake of brevity. Utilizing the AR coefficient estimate ${\widehat{\text{A}}}_{\text{AML}}^{H}$, we obtain the temporally whitened data matrices $\stackrel{\u0304}{\stackrel{\u0303}{\text{S}}}$ and ${\stackrel{\u0304}{\stackrel{\u0303}{\text{x}}}}_{0}$ respectively. Inserting these estimates in (22), we finally come up with the adaptive CGPGLRTAML detector:
In general, the adaptive versions of the CGPGLRT detector can be defined as a parametric implementation of the following adaptive GLRTLQ detector [34]
where the original GLRTLQ detector given in [10, 11] has been reintroduced in (54). The spacetime covariance matrix estimate $\widehat{\text{R}}$ can be also obtained by exploiting the SCM, NSCM, and AML estimators, respectively. However, taking the SCM estimator as an example, we at least need K ≥ JN training signals to ensure a fullrank estimate of the JN × JN matrix R. Obviously, the large JN spatiotemporal product will impose excessive training and computational burdens to the detector.
3.3 Complexity issues
We provide the floatingpoint operations (flops) involved in each step of the parametric and nonparametric STAP detector with different covariance estimators in Tables 1 and 2, respectively. Only the complex multiplication and division are considered but the complex addition and subtraction are ignored.
Note that the adaptive parametric implementations differ mainly in parameter estimation, they share identical steps in signal whitening and calculating the test statistic. The CGPGLRTNSCM detector is slightly more complex than the CGPGLRTSCM detector since it requires evaluating ${\widehat{\tau}}_{{k}_{\text{NSCM}}}$ by utilizing ${\widehat{\text{A}}}_{\text{SCM}}^{H}$ for normalization. However, the CGPGLRTAML detector is the most complex detector among the three. Likewise, similar conclusion can be made for the adaptive versions of the GLRTLQ detector. For a quick comparison, suppose KN > JP for the parametric detectors in Table 1. Clearly, it can be seen from Tables 1 and 2 that the parametric detectors can reduce the computational complexity of their nonparametric counterparts, especially when the spatialtemporal dimension JN is large.
4 Performance assessment
This section is devoted to the performance analysis of the proposed detectors: CGPGLRT, CGPGLRTSCM, CGPGLRTNSCM, and CGPGLRTAML in terms of the probability of false alarm (P_{fa}) and the probability of detection (P_{ d } ). For compoundGaussian environment, the closedform expressions for both P_{fa} as well as P_{ d } are not available. Hence, we carry out the analysis via MonteCarlo techniques based on 100/P_{fa} and 100/P_{ d } independent trials, respectively. In order to limit the computational burden, we set the probability of false alarm P_{fa} = 10^{3} and the number of sensors J = 4 throughout the section.
Moreover, the following statements hold for this section:
(i) The spatial driving noise is Kdistributed, which is obtained assuming that the texture component is Gamma distributed [35]
where Г(·) is the gamma function, μ = E{τ} denotes the mean of the distribution, and ν is the shape parameter which provides a measure of noise spikiness. Lower ν means more spikes will appear in disturbance. Without loss of generality, we set μ = 1.
(ii) The disturbance signal is generated as a multichannel AR(2) compoundGaussian process with the method in [36, 37]. We assume that the order P of the multichannel AR process is known. If P is unknown, it can be selected using some model order selection criterion [38]. The following correlation matrices are utilized to synthesize simulated data [39]
and
where ρ relates to the disturbance power, ρ _{ s } decides the spatial correlation, ρ _{ t } controls the temporal correlation, and {(m  n + l)ω} defines the phase of the correlation function (56). Here, we choose the value of ρ to satisfy tr{Q} = J.
(iii) The target steering vector sequence {s(n)} for a uniform linear array is of the form [15]
where f_{ ts } and f_{ td } denote the target normalized spatial and Doppler frequencies, respectively, and the target spatial steering vector s(f_{ ts } ) is defined as $\text{s}\left({f}_{ts}\right)=\left(1/\sqrt{J}\right){\left[1\phantom{\rule{1em}{0ex}}{e}^{j2\pi {f}_{ts}}\dots {e}^{j2\pi \left(J1\right){f}_{ts}}\right]}^{T}$. In simulation, we set f_{ ts } = 0 and f_{ td } = 0.25.
(iv) The SINR has been defined as
For reader's convenience, we give an overview of relevant parameter settings for the simulation processes in Table 3.
In Figures 1, 2, 3, 4, and 5 the CFAR properties of the proposed detectors: CGPGLRT, CGPGLRTSCM, CGPGLRTNSCM and CGPGLRTAML are investigated. Figure 1 plots the detection threshold versus shape parameter ν for the proposed detectors with training size K = 2 and fixed P_{fa} = 10^{3}. Relevant test parameters are taken from Table 3 and specified in the figures. Observe that the CGPGLRT, CGPGLRTNSCM, and CGPGLRTAML thresholds are almost insensitive to changes in shape parameters ν. This reflects their texture CFAR behavior. However, for ν < 1, the thresholds of CGPGLRTSCM increases with decreasing ν (increasing disturbance spikiness). Thus, a loss of CFAR with respect to the texture PDF comes about for the CGPGLRTSCM detector.
Figures 2, 3, 4, and 5 depict plots of P_{fa} versus threshold corresponding to several shape parameter values (ν = 0.1, 0.5, 1, 4.5), for the proposed CGPGLRT, CGPGLRTSCM, CGPGLRTNSCM, and CGPGLRTAML, respectively. The curves for the CGPGLRTSCM in Figure 3 show much higher variability compared to the CGPGLRT, CGPGLRTNSCM, and CGPGLRTAML, which confirms its lack of CFAR with respect to the texture variations. The P_{fa} plots for the CGPGLRT, CGPGLRTNSCM, and CGPGLRTAML also validate their robust texture CFAR performance.
In Figure 6, the effects of the disturbance spikiness upon the detection performance of the CGPGLRT detector are analyzed by varying the shape parameter ν. Figure 6 displays P_{ d } plots versus SINR for shape parameters ν = 0.1, 0.5, 1, 4.5, 10. The figure highlights that the detection performance is enhanced as ν decreases at the low SINR region. However, at the large SINR region (about > 10 dB), the CGPGLRT detector performs worse with smaller ν. The lower the value of ν and the spikier the disturbance, for a given average disturbance power μ, at low SINR (small value of α), the signal energy that can be accumulated from the cells under test increases, hence, the detection performance is improved; at large SINR (large value of α), more spikes in disturbance may have a negative effect on the signal detection. In addition, the detector exhibits close performance for ν = 4.5 and 10, since the K distribution collapses into the Gaussian distribution when ν ≥ 4.5.
In Section 3.2, we combined the multichannel AR parameter estimation algorithm with three covariance matrix estimation strategies: SCM, NSCM, and AML estimators, and then gave three adaptive CGPGLRT detectors, where the true multichannel AR parameters Q and A^{H} are substituted with the estimated ones. Now various simulations are performed to evaluate the detection performance loss of the adaptive CGPGLRT detectors with respect to the CGPGLRT detector which possesses the perfect knowledge about the multichannel AR parameters Q and A^{H} .
In Figures 7, 8, and 9, the limitedtraining (K = 2) case is considered for different numbers of pulses N (N = 16, 32) and shape parameters ν = 0.1, 0.5 and 4.5. In particular, Figure 7 refers to the shape parameter ν = 0.5 whereas Figures 8 and 9 to ν = 0.1 and ν = 4.5. All the detection curves in Figures 7, 8, and 9 show that the detection performance is improved as N increases. For instance, at P_{ d } = 0.9, the CGPGLRTAML is about 4 dB from the CGPGLRT bound at N = 16; the gap reduces to about 1.5 dB at N = 32 for any value of ν in this case. This implies that we can remedy the performance degradation with respect to CGPGLRT detector by increasing the temporal dimension N. In Figures 7 and 8, we find that the CGPGLRTNSCM bears almost the same performance as the CGPGLRTAML at N = 32 and slightly poorer than the CGPGLRTAML at N = 16. Figure 9 demonstrates that all the adaptive detectors: CGPGLRTSCM, CGPGLRTNSCM, and CGPGLRTAML perform almost the same in the homogeneous environment (ν ≥ 4.5), simply because the K distribution converges to the Gaussian distribution for increasing ν. Thus, the loss of the CGPGLRT detector with the SCM estimator due to model mismatch decreases as ν increases. We also observe that the CGPGLRTAML detector outperforms the other two detectors for the heavy tailed disturbance (ν → 0). Thus, the CGPGLRTAML detector is the most suitable detector to implement the adaptive detection in the realistic spiky disturbance environment.
In Figures 10 and 11, we compare the performance of the proposed adaptive CGPGLRT detectors with that of the GLRTLQ with R_{scm} and with R_{nscm}. Figures 10 and 11 refers to the plots of P_{ d } versus SINR of the adaptive CGPGLRT detectors with training size K = 2, 3 and the GLRTLQ with R_{scm}(R_{nscm}) with K = 128. The large K chosen for the GLRTLQ with estimated R is to ensure a nonsingular estimate of R and obtain performance within 3 dB from the optimum bound. It is observed that, with the limitedtraining (K = 2), the performance of the CGPGLRTSCM is close to that of the GLRTLQ with R_{scm}. And both CGPGLRTNSCM and CGPGLRTAML detectors can produce better detection performance with dramatically less training than the GLRTLQ with R_{scm}. Though all the proposed adaptive CGPGLRT detectors with K = 2 are inferior to the GLRTLQ with R_{nscm} with K = 128, the performance degradation of the CGPGLRTNSCM and CGPGLRTAML detectors can be easily remedied only by increasing the training size to K = 3. Indeed, the parametric detectors are significantly less dependent on training than the covariance matrix based detector.
So far the model order P of the multichannel AR process has been assumed to be known. In the following simulation example, we evaluate the performance of the proposed adaptive CGPGLRT detectors when an model order estimation error occurs. Figure 12 depicts the performance of the proposed adaptive CGPGLRT detectors using an overestimated model order. The curves indicate that some performance loss caused by mismatched model order P happens for N = 16 and K = 3. Since P samples have been discarded in the derivation of the proposed detectors as well as in calculating the test statistics, the detection performance is degraded as model order P is overestimated and the number of pulses N is not large enough.
5 Conclusions
In this article, multichannel signal detection problem in spacetime colored compoundGaussian environment is discussed. By exploiting the structural information about the disturbance spacetime covariance matrix, we model the disturbance signal as a multichannel AR process to ease the training and computational burdens. Modeling the texture as an unknown deterministic parameter, we first derive the CGPGLRT detector under the assumption that the multichannel AR parameters Q and A^{H} are perfectly known. For the twostep GLRT design criterion, we combine the multichannel AR parameter estimation algorithm with three covariance matrix estimation strategies, i.e., sample covariance matrix (SCM), normalized sample covariance matrix (NSCM) and approximate ML (AML) estimator, and then obtain the adaptive versions of the CGPGLRT detector by substituting the true multichannel AR parameters with their estimates. Finally, we show the CFAR behavior and detection performances of the proposed detectors: CGPGLRT, CGPGLRTSCM, CGPGLRTNSCM, and CGPGLRTAML by Monte Carlo trials.
Conclusions of this article are summarized as follows:

The CGPGLRTSCM detector has no texture CFAR property, while the CGPGLRT, CGPGLRTNSCM, and CGPGLRTAML detectors ensure texture CFAR property.

The detection probability of the CGPGLRT detector increases with increasing disturbance spikes (decreasing ν) at low SINR. However, at large SINR, we observe a degradation in performance as ν decreases.

For the limitedtraining case, the detection performance loss of the adaptive CGPGLRT detectors with respect to the CGPGLRT detector can be remedied by increasing temporal dimension N.

The CGPGLRTAML detector has the best detection performance in the heavy tailed disturbance (ν → 0). Thus, the CGPGLRTAML detector is the most suitable detector to implement the adaptive detection in the realistic spiky disturbance.

Compared to the covariance matrix based detector, the proposed modelbased adaptive detectors have significantly decreased the training requirements and the computation complexity.
References
 1.
Klemm R: Principles of SpaceTime Adaptive Processing. Inst Elect Eng London U.K.; 2002.
 2.
Ward J: Spacetime adaptive processing for airborne radar. MIT Lincoln Lab Lexington MA; 1994. Technical report TR1015
 3.
Brennan LE, Reed IS: Theory of adaptive radar. IEEE Trans Aerospace Electron Syst 1973, (2):237252. AES9
 4.
Reed IS, Mallett JD, Brennan LE: Rapid convergence rate in adaptive arrays. IEEE Trans Aerospace Electron Syst 1974, (6):853863. AES10
 5.
Kelly EJ: An adaptive detection algorithm. IEEE Trans Aerospace Electron Syst 1986, 22: 115127.
 6.
Chen WS, Reed IS: A new CFAR detection test for radar. Digital Signal Process 1991, 1: 198214.
 7.
Robey FC, Fuhrmann DR, Kelly EJ, Nitzberg R: A CFAR adaptive matched filter detector. IEEE Trans Aerospace Electron Syst 1992, 28(1):208216.
 8.
Michels JH, Himed B, Rangaswamy M: Performance of STAP tests in Gaussian and compoundGaussian clutter. Digital Signal Process 2000, 10(4):309324.
 9.
He Y, Jian T, Su F, Qu C, Gu X: Novel rangespread target detectors in nonGaussian clutter. IEEE Trans Aerospace Electron Syst 2010, 46(3):13121328.
 10.
Conte E, Lops M, Ricci G: Asymptotically optimum radar detection in compoundGaussian clutter. IEEE Trans Aerospace Electron Syst 1995, 31(2):617625.
 11.
Gini F: Suboptimum coherent radar detection in a mixture of K distributed and Gaussian clutter. IEE ProcRadar Sonar Nav 1997, 144(1):3948.
 12.
Conte E, Maio AD: Distributed target detection in compoundGaussian noise with Rao and Wald tests. IEEE Trans Aerospace Electron Syst 2003, 39(2):568582.
 13.
Jay E, Ovarlez JP, Declercq D, Duvaut P: BORD: bayesian optimum radar detector. Signal Process 2003, 83: 11511162.
 14.
Michels JH, Rangaswamy M, Himed B: Performance of parametric and covariance based STAP tests in compoundGaussian clutter. Digital Signal Process 2002, 12(23):307328.
 15.
Roman JR, Rangaswamy M, Davis DW, Zhang Q, Himed B, Michels JH: Parametric adaptive matched filter for airborne radar applications. IEEE Trans Aerospace Electron Syst 2000, 36(2):677692.
 16.
Sohn KJ, Li H, Himed B: Parametric Rao test for multichannel adaptive signal detection. IEEE Trans Aerospace Electron Syst 2007, 43(3):920933.
 17.
Sohn KJ, Li H, Himed B: Parametric GLRT for multichannel adaptive signal detection. IEEE Trans Signal Process 2007, 55(11):53515360.
 18.
Wang P, Li H, Himed B: A new parametric GLRT for multichannel adaptive signal detection. IEEE Trans Signal Process 2010, 58(1):317325.
 19.
Sohn KJ, Li H, Himed B, Markow JS: Performance of multichannel parametric detectors with MCARM data. In Proceedings of 2007 IET International Conference on Radar Systems. Edinburgh U.K.; 2007:15.
 20.
Wang P, Sohn KJ, H Li, Himed B: Performance evaluation of parametric Rao and GLRT detectors with KASSPER and bistatic data. In Proceedings of 2008 IEEE Radar Conference. Rome Italy; 2008:16.
 21.
Abramovich YI, Spencer NK, Turley M: Timevarying autoregressive (TVAR) models for multiple radar observations. IEEE Trans Signal Process 2007, 55(4):12981311.
 22.
Marple SL Jr, Corbell PM, Rangaswamy M: Multichannel fast parametric algorithms and performance for adaptive radar. In 2007 Conference Record of The FortyFirst Asilomar Conference on Signals, Systems and Computers (ACSSC 2007). Volume 1. Pacific Grove CA; 2007:18351838.
 23.
Marple SL Jr, Corbell PM, Rangaswamy M: Performance tradeoffs for multichannel parametric adaptive radar algorithms. In Proceedings of the 2008 International Conference on Radar. Adelaide SA; 2008:154159.
 24.
Rangaswamy M, Michels JH: A parametric multichannel detection algorithm for correlated nonGaussian random processes. In Proceedings of the 1997 IEEE National RADAR Conference (NATRAD'97). Syracuse NY; 1997:349354.
 25.
Conte E, Maio AD, Ricci G: Covariance matrix estimation for adaptive CFAR detection in compoundGaussian clutter. IEEE Trans Aerospace Electron Syst 2002, 38(2):415426.
 26.
Gini F, Greco M: Covariance matrix estimation for CFAR detection in correlated heavy tailed clutter. Signal Process 2002, 82: 18471859.
 27.
Michels JH, Himed B, Rangaswamy M: Robust STAP detection in a dense signal airborne radar environment. Signal Process 2004, 84(9):16251636.
 28.
Kay SM: Modern Spectral Estimation: Theory and Application. PrenticeHall Englewood Cliffs NJ; 1988.
 29.
Michels JH, Himed B, Rangaswamy M: Evaluations of the normalized parametric adaptive matched filter STAP test in airborne radar clutter. In Proceedings of the International Radar Conference. Alexandria VA; 2000:769774.
 30.
Marple SL Jr: Digital Spectral Analysis With Applications. PrenticeHall Upper Saddle River NJ; 1987.
 31.
Pulsone NB, Raghavan RS: Analysis of an adaptive CFAR detector in nonGaussian interference. IEEE Trans Aerospace Electron Syst 1999, 35(3):903916.
 32.
Wang P, Li H, Himed B: Parametric Rao tests for multichannel adaptive detection in partially homogeneous environment. IEEE Trans Aerospace Electron Syst 2011, 47(3):18501861.
 33.
Greco M, Gini F, Younsi A, Rangaswamy M, Zoubir A: Nonstationary Sea clutter: impact on disturbance covariance matrix estimate and detector CFAR. In Proceedings of the 2008 International Conference on Radar. Adelaide SA; 2008:558562.
 34.
Shuai X, Kong L, Yang J: Performance analysis of GLRTbased adaptive detector for distributed targets in compoundGaussian clutter. Signal Process 2010, 90: 1623.
 35.
Gini F, Greco M: Suboptimum approach for adaptive coherent radar detection in compoundGaussian clutter. IEEE Trans Aerospace Electron Syst 1999, 35(3):10951104.
 36.
Michels JH, Varshney PK, Weiner DD: Synthesis of correlated multichannel random processes. IEEE Trans Signal Process 1994, 42(2):367375.
 37.
Rangaswamy M, Weiner D, Ozturk A: Computer generation of correlated nonGaussian radar clutter. IEEE Trans Aerospace Electron Syst 1995, 31(1):106116.
 38.
Stoica P, Moses RL: Spectral Analysis of Signals. Pearson/PrenticeHall Upper Saddle River NJ; 2005.
 39.
Jiang Y, Stoica P, Li J: Array signal processing in the known waveform and steering vector case. IEEE Trans Signal Process 2004, 52(1):2335.
Acknowledgements
The authors are very grateful to the anonymous referees for their careful reading, helpful comments and constructive suggestions on improving the exposition of this article. This study is supported by the National Natural Science Foundation of China under Grant No. 11074270 and No. 61172166.
Author information
Affiliations
Corresponding author
Additional information
Competing interests
The authors declare that they have no competing interests.
Authors’ original submitted files for images
Below are the links to the authors’ original submitted files for images.
Rights and permissions
Open Access This article is distributed under the terms of the Creative Commons Attribution 2.0 International License ( https://creativecommons.org/licenses/by/2.0 ), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
About this article
Cite this article
Xu, Q., Ma, X., Yan, S. et al. Multichannel adaptive signal detection in spacetime colored compoundgaussian autoregressive processes. EURASIP J. Adv. Signal Process. 2012, 69 (2012). https://doi.org/10.1186/16876180201269
Received:
Accepted:
Published:
Keywords
 compoundgaussian disturbance
 multichannel autoregressive process
 generalized likelihood ratio test (GLRT)
 covariance matrix estimation
 adaptive detection