- Research Article
- Open Access
Oblique Projection Polarization Filtering-Based Interference Suppressions for Radar Sensor Networks
© Bin Cao et al. 2010
- Received: 28 November 2009
- Accepted: 22 March 2010
- Published: 3 May 2010
The interferences coming from the radar members degrade the detection and recognition performance of the radar sensor networks (RSNs) if the waveforms of the radar members are nonorthogonal. In this paper, we analyze the interferences by exploring the polarization information of the electromagnetic (EM) waves. Then, we propose the oblique projection polarization filtering- (OPPF-) based scheme to suppress the interferences while keeping the amplitude and phase of its own return in RSNs, even if the polarized states of the radar members are not orthogonal. We consider the cooperative RSNs environment where the polarization information of each radar member is known to all. The proposed method uses all radar members' polarization information to establish the corresponding filtering operator. The Doppler-shift and its uncertainty are independent of the polarization information, which contributes that the interferences can be suppressed without the utilization of the spatial, the temporal, the frequency, the time-delay and the Doppler-shift information. Theoretical analysis and the mathematical deduction show that the proposed scheme is a valid and simple implementation. Simulation results also demonstrate that this method can obtain a good filtering performance when dealing with the problem of interference suppressions for RSNs.
- Target Signal
- Linear Frequency Modulation
- Oblique Projection
- Interference Suppression
Due to the complex and time-varying environments, also with respect to the manmade or natural interferences, the detection and recognition performance of the single radar system is limited by several issues. Since the techniques on electromagnetic countermeasure (ECM) and counter-counter measures (CCM) are developing more and more sophisticated, we can foresee that it is considerably difficult to fulfil the goal of giving the most accurate interpretation about what the target is at any given point in time for any detection and recognition system. It is also known to all that the slow fluctuations of the target radar cross section (RCS), which result in the radar target fades, are a main factor of degrading the detection and recognition performance of the radar systems .
The single radar systems encountering those problems mentioned above boost the radar researchers and engineers to exploit the new and effective schemes, therein, the radar sensor networks (RSNs) appear [2–10]. The RSNs are promised to operate with multiple goals managed by an intelligent platform network that can manage the dynamics of each radar member to meet the common goals of the platform, rather than each radar to operate as an independent system. This is the so-called cooperative radar networks with communication function . While the performance of RSNs degrades because the radar members are likely to interfere with each other if their waveforms are not orthogonally designed, or their polarized states of the waveforms are not properly allocated . In order to eliminate these interferences, much attention has been paid to the waveform design for RSNs [4–7, 9, 10]. The utilization of the polarization information for RSNs attracts relatively little attention than the waveform design, while our work tries to exploit the polarization information of the electromagnetic (EM) waves for RSNs, and we expect to seek an effective and easy implementation for RSNs to mitigate the interferences based on the use of the polarization information.
For the single radar systems, the polarization filtering (PF) technique is an effective method of interference suppressions, and the PF attracts much attention in recent decades [11–18]. While the conventional polarization filtering (CPF) [11–13] suffers distortions on both amplitude and phase of the target signal when establishing the orthogonal complementary vector of the interference polarization to cancel the interference, because of the neglect of the target polarization. The distortions on target is called the polarization loss introduced by the PF , and the polarization loss is determined by the distance when representing the polarization of the target and the interference in the form of the Poincare-Sphere. If and only if the distance between them is 180°, that is, their polarizations are orthogonal, no polarization loss is introduced. The polarization loss limits the filtering performance of the PF. To avoid the distortions on target, the null-phase-shift polarization filtering (NPSPF) was proposed in . Although the NPSPF can solve the problem of distortions, it would suppress both the interference and the target when they hold the same polarized angle, and the CPF also encounters this problem when the target and the interference are both vertical polarization. These issues limit the application scopes of the PF. Meanwhile, little attention has been paid to the applications of the PF for RSNs.
In this paper, we propose a novel polarization filtering-based approach for RSNs, the suggested scheme suppresses the interferences from the radar members by using the oblique projection polarization filtering (OPPF) [16–20], and the OPPF is the extension to the CPF and the NPSPF. The proposed OPPF can separate the signals effectively if the polarization information of them is different, and the polarized states are not needed to be orthogonal due to the merits of the oblique projections . The detailed implementation of the OPPF is to construct the polarization subspaces of the target and the interference, respectively, then the filtering operator is established according to the oblique projection operator. After passing through the OPPF, the interferences (other radar members' returned signals) are effectively cancelled while keeping the desired return with the same amplitude/phase and the polarization before the operation by the OPPF, even if their waveforms and polarized states are not orthogonal. The proposed scheme can effectively separate the desired returned signal and the interferences without additional transformation and compensation processing. With the desired return not suffering distortions after separation, the scheme is still valid when the desired return and the interferences hold the same polarized angle but with different phase difference in polarized angle.
Since the polarization filtering is independent of the frequency, the spatial, and the temporal domains, the proposed design is expected to exploit more resources which can achieve better suppression performance of the clutter and the interferences for RSNs. The Doppler-shift and its uncertainty are also independent of polarization, and this shows the implementation of the suggested scheme is simpler than the waveform-design-based systems.
We consider that all radar members' polarization information are shared among each radar member, that is, in a cooperative way. The proposed method utilizes all of the radar members' polarization information to establish the corresponding filtering operator for each radar member. Theoretical analysis and the mathematical deduction show that the proposed scheme is a valid and simple implementation. Simulation results also demonstrate that this method can obtain a good filtering performance when dealing with the problem of interference suppressions for RSNs.
The remainder of this paper is organized as follows. System model and fundamentals of the oblique projections are introduced in Section 2. The OPPF-based interference suppressions for the cooperative scenario is discussed in Section 3. Detailed analysis and the simulation results are done and illustrated in Section 4. Finally, Section 5 concludes this paper.
A scalar is represented by any symbol in an italic font, for example, . Vectors are column vectors and are represented by a boldface symbol, such as . Matrices are represented by symbols in a bold font and are usually uppercase, such as . The subspace spanned by the columns of a matrix is represented with angle brackets around the symbol for the matrix, for example, . We use the symbol to denote the orthogonal complement of . The symbol denotes the complex Euclidean space of dimension . We use to represent an -by- array matrix defined in the complex Euclidean space of dimension . is the unit matrix. is the operation of calculating the expectation.
A superscript is used to indicate the transpose of a matrix or vector, such as , and denotes the Hermitian transpose, for example, . is the dot product of two vectors, is the symbol of pseudoinverse of a matrix, such as , and denotes the conjugate operating. is the imaginary part unit, and is the conjugate of a complex number.
2.1. System Model
If , the polarization is linear.
, indicates the right-handed circular polarization, and , shows the left-handed circular polarization.
If and , then the polarization is left-handed elliptic. When and , the polarization is right-handed elliptic.
respectively, where indicates the argument of the vector . After being radiated by the ODPA, the EM waves may produce the so-called cross-polarization (XP) component whose polarization is orthogonal to the original transmitted polarization, and this is known as the effect of depolarization [21, 22]. The degree of the depolarization is described by the cross-polarization discrimination (XPD) which indicates the ratio of the original transmitted polarization component power and the cross-polarization component power at the same location point. The depolarization effect can be eliminated by introducing the depolarization compensation technology .
There are also many methods for estimation of the polarized states (see [15, 24]). In a communication realization, the polarized states can be estimated in the frequency domain if each user holds different frequency bands . If they overlap in the frequency domain, then we can estimate their polarized states in the Fractional Fourier domain, because the modulation rate or the starting frequency of each member may be different. By using the Fractional Fourier Transformation (FrFT), their polarized states can be estimated effectively. We assume that the polarized state is already known at the receiver in this paper.
where is the amplitude of the return of the th radar member; and are the polarization information of its own return. If the target is a fluctuating one, the most popular and reasonable model for is the "Swerling 2" model. Therefore, follows the Rayleigh distribution .
The polarized state suffers changes during the propagation [21, 22], for example, the reflections and the shadowing effect can destroy the polarization. Hence, the estimation of the polarized states (or polarization measurement) is essential for the PF. When the polarized states suffers changes, the depolarization compensation technology is valid to compensate the changes .
where is the amplitude of the returned signal from the th radar member; is the additive white Gaussian noise (AWGN) with mean value zero and variance .
In order to suppress those components coming from other radar members, several work based on the waveform design of the signal has been proposed [4–7, 9, 10]. The propagation processing also destroys the orthogonality of the waveforms. Due to the movement of the target, the relative speed rate of the target to each radar member is different, and the Doppler-shift and its uncertainty are essential to the waveform-design-based systems. In this paper, we make use of the polarization information to suppress those interferences. It is obvious that the polarization is independent of the frequency, and the Doppler-shift and its uncertainty.
2.2. Oblique Projections
Herein, we consider two full column rank matrices and , and suppose . The columns of and are nonoverlapped indicating that the intersection of the range subspaces and , respectively, spanned by and only contains vector . It is obvious that the condition of nonoverlapped implies that the vector spaces are linear independent. If their intersection of the range subspaces only contains vector , then they are called as disjointed vector spaces. Two vectors are disjointed does not imply that they are orthogonal with each other. It is easy to find that the composite matrix is also a full column rank matrix with rank .
It can be found that the range space of the oblique projection operator is , and is a subset of its null subspace.
As an extension to the orthogonal projection, oblique projection does not need the condition that the two subspaces are orthogonal. If they are orthogonal, the projection operator becomes the orthogonal projection operator.
We consider that all radar members share their polarization information. For convenience, we assume that there are two radar members in the RSN, that is, . It is proved that the OPPF can solve the problem of interference suppressions when there are more than two radar members . We propose the polarization vector transformation (PVT) to fulfil the multiinterference suppressions based on the merits of the oblique projections.
Obviously, and are column vectors and are both full column rank (rank is 1). If the polarized states of the target and the interference are different, then the subspaces and are disjointed, where the composite matrix is a full column rank matrix.
where is the orthogonal projection operator onto the interference polarization subspace .
where is the result after passing through .
Formula (15) demonstrates that the interference is suppressed totally, and the output is the original target signal plus the additive Gaussian noise, with the amplitude and phase of the target signal remaining the same after projection processing. As long as the polarized states of the target signal and the interference are different, that is, the polarized angles and the phase difference in polarized angle are not equal simultaneously, then the oblique projection operator can suppress the interference totally and keep the amplitude and phase information of the target. The proposed scheme does not require the information of the frequency, the time-scale, and the Doppler-shift and its uncertainty. If the target signal and the interference are overlapped in the frequency domain, the interference can also be suppressed by using the oblique projection operator.
The OPPF operator contains the polarization information of the target and the interference. If the polarized states of both the target signal and the interference are known, the OPPF operator can be constructed according to the formulaes (18)-(19) to extract the target signal.
Along the same analysis flow, it can be found out that the oblique projection operator along subspace onto subspace can also extract the components of . Hence, in the cooperative RSN, each radar can obtain its own target signal while suppressing other radar members' interfering signals, since and .
Consider that there are more than two radar members in the RSN, that is, . For simplicity of analysis, we assume there are three radar members, while it can be extended to more radar members in a straightforward way.
It is easy to find that the linear model for oblique projections is not a full column rank matrix, the model is a matrix belonging to , that is, . In order to construct a model that satisfies the full-rank property, we utilize the polarization vector transformation (PVT) to transform the target signal into a vertically polarized waves.
Let , and , that is, denotes the vertical polarization. It is easy to obtain that . Then after the transformation, the returned signal of the 1st user is a vertically polarized waves.
where , , , and ( ) are the parameters after transformation by .
where , and are both real numbers. Particularly, and can be chosen as the modulus of the complex number and , or can be also just selected the real part of and .
After the projection process, the other two radar members' returned signals are suppressed. We select only the first two array components of the projection results since the third array is a virtual component.
For more than three radar members, we can select and as different parameters. Take four radar members as an instance, for the st user, it should add at least two virtual components to its true received signals. The first virtual component can be selected according to the analysis mentioned above, and the second component can be selected according to the following principle: transform the second radar member's polarization to vertical polarization, then repeat the same process.
In order to fully understand the performance of the proposed OPPF, the detailed analysis and the simulation results are presented in this section.
If the polarized states are not exactly estimated, that is, some estimation errors are introduced, we also consider the case of two-radar-member in the RSN for simplifying the complexity of analysis. For the 1st user, if cannot be estimated exactly, will suffer distortions after passing through the OPPF, since the oblique projection cannot be established correctly. And if the estimation of is not the exact value, the component of cannot be suppressed totally.
When there is estimation error on , the interference cannot be totally suppressed for the st radar member. The performance degradation brought by the estimation error is analyzed by the relationship between the error deviation on and the gain of signal-to-interference ratio .
By (38), the gain of signal-to-interference ratio has a relationship with the difference between and the error deviation . The larger the is, the larger the can be obtained. In order to get good performance, a moderately large difference between the polarized angles of each radar member should be designed.
Fig. 4 describes the relationship between , and . When is 0.1° and is 60°, he gain of signal-to-interference ratio which can be obtained is about 54dB, and if is 30°, the gain of signal-to-interference ratio is 50 dB. When is 1° and are 60° and 45°, the gains of signal-to-interference ratio are 34 dB and 32 dB, respectively. This indicates that the gain of signal-to-interference ratio is sensitive to .
To show the performance when the RSN adopts the proposed OPPF intuitively, we verify its performance by using simulations.
These simulation results show that once the polarization of each radar member is different from each other, the proposed OPPF are valid when dealing with the problem of interference suppressions for RSNs. Moreover, after the operation of the OPPF, the diversity combination can also be available, besides the spatial and the time diversity, the polarization diversity can also be exploited in the OPPF-based RSNs.
In the RSNs, the radar members interfere with each other if their waveforms are nonorthogonal and this introduces degradation to the detection and estimation performance of the RSNs. In this paper, we introduce the polarization information rather than the waveform design for RSNs, and discuss the feasibility when using the proposed oblique projection polarization filtering (OPPF)-based scheme to suppress the interferences in a cooperative RSN where all the polarization information of the radar member are shared. The analysis shows that the Doppler-shift and its uncertainty are independent of the polarization information, which contributes that the interferences can be suppressed without the utilization of the spatial, the temporal, the frequency, the time-delay, and the Doppler-shift information. Theoretical analysis and the mathematical deduction show that the proposed scheme is a valid and simple implementation to the interference suppressions problems for RSNs. Simulation results also illustrates a good filtering performance when adopting the proposed OPPF for the cooperative RSN.
This work has been supported by the National Natural Sciences Foundation of China (NSFC) under the Grant No. 60432040 No. 60702034 and the National Basic Research Program of China under Grant No. 2007CB310606.
- Skolnik M: Introduction to Radar Systems. 3rd edition. McGraw Hill, New York, NY, USA; 2001.Google Scholar
- Łubczonek J, Stateczny A: Aspects of spatial planning of radar sensor network for inland waterways surveillance. Proceedings of the 6th European Radar Conference (EuRAD '09), 2009 501-504.Google Scholar
- Liang Q: Collaborative signal processing using radar sensor networks. Proceedings of the IEEE Military Communications Conference (MILCOM '06), 2006 1-6.Google Scholar
- Xu L, Liang Q: Radar sensor network using a new triphase coded waveform: theory and application. Proceedings of the IEEE International Conference on Communications (ICC '09), 2009 1-5.Google Scholar
- Liang Q: Radar sensor networks for automatic target recognition with delay-doppler uncertainty. Proceedings of the IEEE Military Communications Conference (MILCOM '07), 2007 1-7.Google Scholar
- Liang J, Liang Q, Zhou Z: Radar sensor network design and optimization for blind speed alleviation. Proceedings of the IEEE Wireless Communications and Networking Conference (WCNC '07), 2007 2645-2649.Google Scholar
- Liang Q: Waveform design and diversity in radar sensor networks: theoretical analysis and application to automatic target recognition. Proceedings of the 3rd Annual IEEE Communications Society on Sensor and Ad Hoc Communications and Networks (Secon '06), 2006 2: 684-689.View ArticleGoogle Scholar
- Pescosolido L, Barbarossa S, Scutari G: Radar sensor networks with distributed detection capabilities. Proceedings of the IEEE Radar Conference (RADAR '08), 2008 1-6.Google Scholar
- Liang J, Liang Q: Orthogonal waveform design and performance analysis in radar sensor networks. Proceedings of the IEEE Military Communications Conference (MILCOM '06), 2006 1-6.Google Scholar
- Ly HD, Liang Q: Spatial-temporal-frequency diversity in radar sensor networks. Proceedings of the IEEE Military Communications Conference (MILCOM '07), 2007 1-7.Google Scholar
- Poelman AJ: Virtual polarisation adaptation-a method of increasing the detection capability of a radar system through polarisation-vector processing. IEE Proceedings. Part F 1981, 128(5):261-270.Google Scholar
- Gherardelli M, Giuli D, Fossi M: Suboptimum adaptive polarisation cancellers for dual-polarisation radars. IEE Proceedings, Part F 1988, 135(1):60-72.Google Scholar
- Poelman AJ: Virtual polarization adaption—a method of increasing detection capability of a radar system through polarization-vector processing. IEE Proceedings, Part F 1981, 128(5):261-270.Google Scholar
- Mao X, Liu Y: Validity of polarization filtering technique. Journal of Harbin Institute of Technology 2002, 34(4):577-580.Google Scholar
- Mao X-P, Liu Y-T: Null phase-shift polarization filtering for high-frequency radar. IEEE Transactions on Aerospace and Electronic Systems 2007, 43(4):1397-1408.View ArticleGoogle Scholar
- Cao B, Liu A-J, Mao X-P, Zhang Q-Y: An oblique projection polarization filter. Proceeding of the International Conference on Wireless Communications, Networking and Mobile Computing (WiCOM '08), 2008 1893-1896.Google Scholar
- Cao B, Zhang Q, et al.: Blind adaptive polarization filtering based on oblique projection. Proceedings of the IEEE International Conference on Communications (ICC '10), 2010Google Scholar
- Wang J, Zhang Q-Y, Cao B: Multi-notch polarization filtering based on oblique projection. Proceedings of the Global Mobile Congress (GMC '09), 2009 1-5.Google Scholar
- Zhang Q, Cao B, Wang J, Zhang N: Polarization filtering technique based on oblique projections. Science in China, Series F 2010, 2010(53):1056-1066.MathSciNetGoogle Scholar
- Behrens RT, Scharf LL: Signal processing applications of oblique projection operators. IEEE Transactions on Signal Processing 1994, 42(6):1413-1424. 10.1109/78.286957View ArticleGoogle Scholar
- Huang J, Wang Y: Application of Weibull distribution in prediction for the rain induced depolarization of millimeter waves. Acta Electronica Sinica 1993, 21(12):93-96.Google Scholar
- Yang R, Huang J, Lv X: Identification for the rain induced depolarization discrimination in millimeter wave propagation. Journal of Xidian University (Natural Science) 2000, 27(4):487-490.Google Scholar
- Zheng L, Liu G: Adaptive compensation techniques for rain depolarization at millimeter wavelength. Journal of Xidian University (Natural Science) 1989, 16(2):185-196.MathSciNetGoogle Scholar
- Roueff A, Chanussot J, Mars JI: Estimation of polarization parameters using time-frequency representations and its application to waves separation. Signal Processing 2006, 86(12):3714-3731. 10.1016/j.sigpro.2006.03.019MATHView ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.