Table of Contents Author Guidelines Submit a Manuscript
Mathematical Problems in Engineering
Volume 2019, Article ID 7010389, 6 pages
https://doi.org/10.1155/2019/7010389
Research Article

Direction Finding of Coherent Sources Using a MIMO Array of Triaxial Velocity Sensors

Air and Missile Defense College, Air Force Engineering University, Xi’an 710051, China

Correspondence should be addressed to Guimei Zheng; moc.361@mg-gnehz

Received 21 August 2019; Accepted 11 October 2019; Published 11 November 2019

Academic Editor: Raffaele Solimene

Copyright © 2019 Yuwei Song et al. 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.

Abstract

In the received side, triaxial velocity sensors of MIMO array are used to solve the problem of coherent source direction-finding in this paper. A new velocity field smoothing algorithm is presented to decorrelate coherent sources. The identically oriented velocity sensors of whole array are divided into three subarrays. Then, the covariance matrices of the three subarrays are smoothed to restore the rank of source covariance matrix (SCM). Lastly, the cross-correlation coefficients of the SCM after smoothing processing are calculated to analyze the performance of decorrelation. The proposed decorrelation algorithm (1) does not need the information of locations of velocity vector sensors; (2) is suitable for arbitrary configuration array; and (3) has no loss of array effective aperture. Simulation results prove the effectiveness of the proposed algorithm.

1. Introduction

Traditional parameter estimation techniques with array signal are based on a scalar sensor array. In order to improve parameter estimation accuracy, the acoustic vector sensor is proposed [1, 2] (it can be also the electromagnetic vector sensor). An acoustic vector sensor comprises two or three orthogonally collocated velocity sensors and a pressure sensor. Compared with the conventional scalar sensor, the vector sensor has more output information and provides more manners of signal processing.

An estimation of signal parameters via rotational invariance techniques- (ESPRIT-) based algorithm is investigated for arbitrarily spaced three-dimensional arrays of vector hydrophones to estimate the two-dimensional direction of arrival (DOA) [3]. But the array aperture is not fully utilized. Aperture extension is achieved for a uniform rectangular array of vector hydrophones spaced much farther apart than a half-wavelength by using a two-step ESPRIT-based method [4]. The same idea of [4] can be applied to the identical subarrays on a sparse uniform rectangular array, but not limited to the acoustic vector sensor [5]. Multiple signal classification (MUSIC) is another popular super-resolution algorithm. A self-initiating MUSIC-based algorithm for the acoustic vector sensor is proposed, which does not require the initial value of the MUSIC algorithm by using beamspace information [6]. Then, the Root-MUSIC-based algorithm without the angle search is studied to reduce the complexity [7]. The above results [37] assume that all signals impinging from far-field. Therefore, an ESPRIT-based algorithm for the near-field is addressed by using a single vector hydrophone [8]. A direction-finding and blind interference rejection algorithm for acoustic vector sensor array is presented, which can solve up to three fast frequency-hop wideband signals [9]. Although these papers have made outstanding contributions, they are all based on the acoustic vector sensors with spatially collocated in a point-like geometry. Therefore, the spatially spread vector-sensor is proposed [10, 11], which not only retains the advantages of the collocated acoustic vector sensors, but also will significantly extend the spatial aperture to improve the direction-finding accuracy by orders of magnitude without additional sensors.

On the other hand, the Multiple-Input Multiple-Output (MIMO) concept was first introduced for radar in 2004, [12, 13]. MIMO radars transmit multiple waveforms and receive signals at multiple antennas. Waveform diversity in colocated MIMO radar enables significant superiority over its phased-array counterpart, including much improved parameter identifiability and estimation accuracy [14]. MIMO radar with widely separated sensors is also a high performance radar [15]. Virtual sensors which can extend array aperture can be obtained by using spatially orthogonal signal transmission. These virtual sensors provide higher performance in target detection, angle estimation accuracy, and angular resolution [16]. In this paper, we study the DOA estimation in colocated MIMO array radar or sonar.

The ESPRIT algorithm is used to estimate the direction of departure (DOD) and DOA by using the invariance property of both the transmit and receive array in a bistatic MIMO radar [17], but it needs an extra pairing processing. The unitary ESPRIT algorithm with real-valued processing can provide better estimation performance than that of ESPRIT with reduced computational complexity and automatic pairing [18]. In order to achieve the signal-to-noise ratio (SNR) gain, transmit beamspace processing is proposed in DOA estimation of MIMO [19].

Combined with the velocity vector sensor and waveform diversity offered by MMO radar, the signal model of the MIMO velocity vector array is given [2022], and then the MUSIC [20] and ESPRIT [21, 22] angle estimation method for the signal model are developed. The results prove that these additional degrees of freedom can enhance spatial resolution, strengthen parameter identifiability, and improve target detection performance.

Although they made outstanding contributions, they are based on the assumption that signals are uncorrected. In practice, coherent signals are often encountered in multipath or some other scenarios, since multipath signals arise from different propagation paths of the same target. For coherent sources direction-finding, the spatial smoothing (SS) algorithm can be used to restore the rank of the covariance matrix. The disadvantage of the SS algorithm is that it will decrease the array aperture. Therefore, the “velocity-field smoothing” (VFS) algorithm restores the rank of the covariance matrix proposed in [2325], whose algorithm will not decrease the array aperture.

In this paper, we extend the VFS algorithm into the MIMO array to solve the coherent sources estimation problem and analyze the performance of decorrelation. The detailed operations of decorrelation method are dividing the identically oriented sensors within all velocity vectors into three identical subarrays and then smoothing the covariance matrices of the three subarrays to restore the rank of source covariance matrix (SCM). The main idea of this paper is to find the identical subarrays in the vector sensor MIMO array. The remainder of the paper is organized as follows. Signal model is addressed in Section 2. In Section 3, we describe the proposed VFS algorithm in MIMO array. In Section 4, performance of the proposed method is evaluated. Section 5 gives the extensive simulations, and Section 6 concludes the paper.

Notation: Superscript , , and denote complex conjugation, transpose, and conjugate transpose, respectively. denotes Kronecker product. denotes identity matrix. denotes the diagonalization of the entity inside. denotes the rank of matrix .

2. Signal Model

We consider a monostatic MIMO sonar with transmitted scalar sensors and triaxial received velocitysensors. Assume that the sensors of transmitted and received array are located in arbitrarily three-dimensional positions. Each of the received sensors consists of three identical but orthogonally oriented velocity sensors, aligning along with the -axis, -axis, and -axis, respectively. The location of the th sensor in the transmitted array is defined as , and the location of the th sensor in received array is defined as . The transmitted sensors transmit orthogonal waveform signals. In each received sensor, the echoes are matched by transmitted waveforms. Assume that coherent signals are located at the far field. After matched filtering processing, the whole output in received side can be written as [22]wherein whichdenote the transmit steering vector and the receive steering vector, respectively. is the propagation vector. and are the two-dimensional DOA, namely, azimuth and elevation, respectively. is defined as wave number. represents the transmitted signals, where , and and represent amplitude and Doppler frequency of the signal, respectively. The sources are set as coherent. The noise is assumed to be white noise with zero mean and is covariance. The matrices , and are equal to

About equation (1), we give more explanations so that readers can understand it better. The data , and are the output of matched filtering using the received data in axes, respectively. Our system is active sonar, but our channel model is very simplistic with line of sight, no diffraction, and only additive spatio-temporally white Gaussian noise. This paper is our preliminary results.

The source covariance matrix will be rank-deficiency because of the coherent of sources. Therefore, the traditional super-resolution algorithms, such as MUSIC and ESPRIT, will be failed. In the following, we proposed a new VFS algorithm to restore the rank of source covariance matrix (SCM).

3. “Velocity-Field Smoothing” Algorithm for MIMO Array

It can be seen from equation (1) that the received data , and in axes have identical form except for the items . Therefore, we can use weighted average of the different items to complete decorrelation for coherent sources. This method is called as VFS algorithm because the velocity field is used in the smoothing processing. In fact, the smoothing processing can be seen as a directly transplantation of spatial smoothing algorithm from spatial domain to velocity-field domain. To make more clear to readers, the following describes VFS smoothing processing with mathematical formulas.

We firstly compute the covariance matrix of the received data of axes separately, and then perform smoothing processing as follows:where . It is well known that the rank of steering matrix equals the number of sources . The sources are coherent, so we can set , where denotes nonzero complex constant. Then, the SCM is equal to , where denotes the power of and the vector of equals . Therefore, the rank of equals 1. Thus, the traditional super-resolution algorithms are failed. But after the smoothing processing, the rank of will restore to the number of sources . In the following, the derivation of the conclusion will be demonstrated. Firstly, can be rewritten aswhere . Note that matrix is a diagonal matrix; therefore, . Recall equation (4), we know that and are linearly independent for arbitrary according to the results of [2325]. Therefore, , and then . From the above analysis, one has . Now, when the number of sources , the rank of SCM will be restored. Therefore, the number of coherent signals can be resolved up to three by using the proposed algorithm.

Remark 1. This algorithm can only resolve up to three coherent sources. The resolvable of coherent sources will be increased to six when forward and backward smoothing technology is used.

Remark 2. It is worth noting that the array structure information is not used, so this proposed method is applicable to arbitrary configuration array.

4. Performance Analysis

It can be seen clearly that the performance of decorrelation depends on the value of , we call it as after-smoothing source covariance matrix (AS-SCM). It is easy to calculate the diagonal elements of , which are equal to , . The nondiagonal element values of AS-SCM are equal to

It can be seen from equation (7) that is a function of two-dimensional DOA. It shows that the performance of decorrelation of the proposed VFS method depends on the parameters of two-dimensional DOA. The smaller the AS-SCM , the better the performance of decorrelation will be. If , the coherent signals will be decorrelated completely.

In order to make the reader more clearly, we give two examples to observe the elements of matrix . Firstly, assume that there are two coherent sources, which are located in and , respectively. Then, we have

Then, the rank of equals . When an extra source with is added, we have

Then, the rank of equals .

5. Computer Simulations

In this section, simulations are used to demonstrate the effectiveness of the proposed algorithm. In the following simulations, we set that the transmitted sensors are equal to and received velocity vector sensors are equal to .

We firstly verify the effectiveness of decorrelation by using eigenvalues of eigendecomposition for AS-SCM. Assume that there are two spatially closed coherent sources with the same power, and sources’ angles are and , respectively. We set that snapshots and SNR = 20 dB. In order to verify that the proposed algorithm is suitable for arbitrary array configuration, the received velocity vector sensors are assumed in the -axis with and the transmitted scalar sensors located in the -axis with . It is clear that the array configuration has no rotational invariance. Therefore, the spatial smoothing algorithm is failed for the above array. For convenient comparison, Figure 1 gives the largest eigenvalues of SCM and AS-SCM. We can clearly see from the figure that there is only one large eigenvalue without smoothing processing. There are two large eigenvalues using VFS smoothing processing, which has the same number to the number of coherent sources. The result demonstrates that the VFS smoothing processing can recover the loss rank of SCM and the proposed algorithm is suitable for arbitrary array configuration. Overall, the results verify the effectiveness of the proposed VFS smoothing algorithm.

Figure 1: Eigenvalues of covariance matrix with and without VFS processing.

In the second simulation, we prove that the performance of our proposed smoothing method depends on the angle of DOA. Without loss of generality, we set . Then, the nondiagonal elements values of AS-SCM are simplified to . Figure 2 shows the nondiagonal elements of AS-SCM versus angle of DOA. It can be seen from the figure that the values of nondiagonal elements are changed with the angle of DOA. The different values of nondiagonal elements values of AS-SCM demonstrate the different performances of the proposed algorithm. A graceful mathematical relationship between decorrelated performance and angle of DOA will be focus of our future work.

Figure 2: The nondiagonal elements of AS-SCM versus DOA.

In the last simulation, we use the eigenvalue distribution and RMSE of DOA estimation to verify the effectiveness of the algorithm further. Root mean square error (RMSE) of azimuth estimation is defined as , where is the number of Monte Carlo experiments, and denote azimuth estimation and real azimuth value of th Monte Carlo experiment with th source, respectively. RMSE of elevation estimation is defined the same as that of azimuth. Here we have two groups’ signals: and . Two groups’ signals mean that the same MATLAB program needs to run two times, and then the two results are compared to prove the performance of decorrelation. According to equation (7), the nondiagonal elements of of the first group signal equals and the second group equals . It implies that the performance of decorrelation of the second group is much better than that of the first group. The same result can be expected in the following figures. Snapshots are equal to . The number of Monte Carlo experiment equals 1000. Assume that the received velocity vector sensors are located in the -axis with half-wavelength uniform linear array and the transmitted scalar sensors are located in the -axis also with half-wavelength uniform linear array. The ESPRIT algorithm is used to estimate sources’ azimuth and elevation after velocity-field smoothing. Figure 3 shows the eigenvalue distribution of AS-SCM of two group signals. It is indicated that the two eigenvalues of the second group are closer to each other than those of the second group, so the performance of decorrelation of the second group is better. The above analysis results are successfully verified. Figure 4 shows the RMSE of two-dimensional DOA estimation along with SNR. We can see from Figure 4 that the DOA estimation performance of the second group is better than those of the first group. It is indicated that the greater the performance of decorrelation, the better estimation accuracy will be.

Figure 3: The eigenvalue distribution of AS-SCM with SNR = 20 dB.
Figure 4: RMSE of two-dimensional DOA estimation versus SNR.

6. Conclusions

This paper has studied the coherent source DOA estimation for MIMO array with received velocity vector sensors. The proposed VFS preprocessing can effectively restore the rank of AS-SCM. The performance of the algorithm is analyzed with computing the nondiagonal elements of AS-SCM. We find that (1) it is a function of DOA and (2) the better smaller the correlation, the better the estimation accuracy will be. In addition, the algorithm is suitable for arbitrary array configuration.

Data Availability

The MATLAB code data used to support the findings of this study are included within the supplementary information file.

Conflicts of Interest

The authors declare that they have no conflicts of interest.

Acknowledgments

The manuscript is supported by the National Natural Science Fund with grant no. 61971438.

Supplementary Materials

The code of velocity-field smoothing algorithm for MIMO array. (Supplementary Materials)

References

  1. A. Nehorai and E. Paldi, “Acoustic vector-sensor array processing,” IEEE Transactions on Signal Processing, vol. 42, no. 9, pp. 2481–2491, 1994. View at Publisher · View at Google Scholar · View at Scopus
  2. M. Hawkes and A. Nehrai, “Acoustic vector-sensor processing in the presence of a reflecting boundary,” IEEE Transactions on Signal Processing, vol. 48, no. 11, pp. 2981–2993, 2000. View at Publisher · View at Google Scholar · View at Scopus
  3. K. T. Wong and M. D. Zoltowski, “Closed-form underwater acoustic direction-finding with arbitrarily spaced vector hydrophones at unknown locations,” IEEE Journal of Oceanic Engineering, vol. 22, no. 3, pp. 566–575, 1997. View at Publisher · View at Google Scholar · View at Scopus
  4. K. T. Wong and M. D. Zoltowski, “Extended-aperture underwater acoustic multisource azimuth/elevation direction-finding using uniformly but sparsely spaced vector hydrophones,” IEEE Journal of Oceanic Engineering, vol. 22, no. 4, pp. 659–672, 1997. View at Publisher · View at Google Scholar · View at Scopus
  5. M. D. Zoltowski and K. T. Wong, “Closed-form eigenstructure-based direction finding using arbitrary but identical subarrays on a sparse uniform Cartesian array grid,” IEEE Transactions on Signal Processing, vol. 48, no. 8, pp. 2205–2210, 2000. View at Publisher · View at Google Scholar · View at Scopus
  6. K. T. Wong and M. D. Zoltowski, “Self-initiating MUSIC-based direction finding in underwater acoustic particle velocity-field beamspace,” IEEE Journal of Oceanic Engineering, vol. 25, no. 2, pp. 262–273, 2000. View at Publisher · View at Google Scholar · View at Scopus
  7. K. T. Wong and M. D. Zoltowski, “Root-MUSIC-based azimuth-elevation angle-of-arrival estimation with uniformly spaced but arbitrarily oriented velocity hydrophones,” IEEE Transactions on Signal Processing, vol. 47, no. 12, pp. 3250–3260, 1999. View at Publisher · View at Google Scholar · View at Scopus
  8. P. Tichavsky, K. T. Wong, and M. D. Zoltowski, “Near-field/far-field azimuth and elevation angle estimation using a single vector hydrophone,” IEEE Transactions on Signal Processing, vol. 49, no. 11, pp. 2498–2510, 2001. View at Publisher · View at Google Scholar · View at Scopus
  9. K. T. Wong, “Acoustic vector-sensor FFH “blind” beamforming & geolocation,” IEEE Transactions on Aerospace and Electronic Systems, vol. 46, no. 1, pp. 444–448, 2010. View at Publisher · View at Google Scholar · View at Scopus
  10. Y. Song and K. T. Wong, “Azimuth-elevation direction finding using a microphone and three orthogonal velocity sensors as a non-collocated subarray,” The Journal of the Acoustical Society of America, vol. 133, no. 4, pp. 1987–1995, 2013. View at Publisher · View at Google Scholar · View at Scopus
  11. Y. Song and K. T. Wong, “Acoustic direction finding using a spatially spread tri-axial velocity sensor,” IEEE Transactions on Aerospace and Electronic Systems, vol. 51, no. 2, pp. 834–842, 2015. View at Publisher · View at Google Scholar · View at Scopus
  12. E. Fishler, A. Haimovich, R. Blum et al., “MIMO radar: an idea whose time has come,” in Proceedings of the 2004 IEEE Radar Conference, vol. 2, pp. 71–78, Philadelphia, PA, USA, April 2004. View at Publisher · View at Google Scholar
  13. E. Fishler, A. Haimovich, R. S. Blum, L. J. Cimini, D. Chizhik, and R. A. Valenzuela, “Spatial diversity in radars-models and detection performance,” IEEE Transactions on Signal Processing, vol. 54, no. 3, pp. 823–838, 2006. View at Publisher · View at Google Scholar · View at Scopus
  14. J. Li and P. Stoica, “MIMO radar with colocated antennas,” IEEE Signal Processing Magazine, vol. 24, no. 5, pp. 106–114, 2007. View at Publisher · View at Google Scholar · View at Scopus
  15. A. Haimovich, R. Blum, and L. Cimini, “MIMO radar with widely separated antennas,” IEEE Signal Processing Magazine, vol. 25, no. 1, pp. 116–129, 2008. View at Publisher · View at Google Scholar · View at Scopus
  16. I. Bekkerman and J. Tabrikian, “Target detection and localization using MIMO radars and sonars,” IEEE Transactions on Signal Processing, vol. 54, no. 10, pp. 3873–3883, 2006. View at Publisher · View at Google Scholar · View at Scopus
  17. C. Duofang, C. Baixiao, and Q. Guodong, “Angle estimation using ESPRIT in MIMO radar,” Electronics Letters, vol. 44, no. 12, pp. 770-771, 2008. View at Publisher · View at Google Scholar · View at Scopus
  18. G. Zheng, B. Chen, and M. Yang, “Unitary ESPRIT algorithm for bistatic MIMO radar,” Electronics Letters, vol. 48, no. 3, pp. 179–181, 2012. View at Publisher · View at Google Scholar · View at Scopus
  19. A. Khabbazibasmenj, A. Hassanien, S. A. Vorobyov, and M. W. Morency, “Efficient transmit beamspace design for search-free based DOA estimation in MIMO radar,” IEEE Transactions on Signal Processing, vol. 62, no. 6, pp. 1490–1500, 2014. View at Publisher · View at Google Scholar · View at Scopus
  20. J. He, M. N. S. Swamy, and M. O. Ahmad, “Joint DOD and DOA estimation for MIMO array with velocity receive sensors,” IEEE Signal Processing Letters, vol. 18, no. 7, pp. 399–402, 2011. View at Publisher · View at Google Scholar · View at Scopus
  21. J. Li and X. Zhang, “Improved joint DOD and DOA estimation for MIMO array with velocity receive sensors,” IEEE Signal Processing Letters, vol. 18, no. 12, pp. 717–720, 2011. View at Publisher · View at Google Scholar · View at Scopus
  22. J. Li and X. Zhang, “Two-dimensional angle estimation for monostatic MIMO arbitrary array with velocity receive sensors and unknown locations,” Digital Signal Processing, vol. 24, no. 24, pp. 34–41, 2014. View at Publisher · View at Google Scholar · View at Scopus
  23. J. Tao, W. Chang, and W. Cui, “Vector field smoothing for DOA estimation of coherent underwater acoustic signals in presence of a reflecting boundary,” IEEE Sensors Journal, vol. 7, no. 8, pp. 1152–1158, 2007. View at Publisher · View at Google Scholar · View at Scopus
  24. J. Tao, W. Chang, and Y. Shi, “Direction-finding of coherent sources via ‘particle-velocity-field smoothing’,” IET Radar, Sonar & Navigation, vol. 2, no. 2, pp. 127–134, 2008. View at Publisher · View at Google Scholar · View at Scopus
  25. J. He, S. Jiang, J. Wang et al., “Particle-velocity-field difference smoothing for coherent source localization in spatially nonuniform noise,” IEEE Journal of Oceanic Engineering, vol. 35, no. 1, pp. 113–119, 2010. View at Publisher · View at Google Scholar · View at Scopus