Research Article  Open Access
Fast Estimation Method of SpaceTime TwoDimensional Positioning Parameters Based on Hadamard Product
Abstract
The estimation speed of positioning parameters determines the effectiveness of the positioning system. The time of arrival (TOA) and direction of arrival (DOA) parameters can be estimated by the spacetime twodimensional multiple signal classification (2DMUSIC) algorithm for array antenna. However, this algorithm needs much time to complete the twodimensional pseudo spectral peak search, which makes it difficult to apply in practice. Aiming at solving this problem, a fast estimation method of spacetime twodimensional positioning parameters based on Hadamard product is proposed in orthogonal frequency division multiplexing (OFDM) system, and the CramerRao bound (CRB) is also presented. Firstly, according to the channel frequency domain response vector of each array, the channel frequency domain estimation vector is constructed using the Hadamard product form containing location information. Then, the autocorrelation matrix of the channel response vector for the extended array element in frequency domain and the noise subspace are calculated successively. Finally, by combining the closedform solution and parameter pairing, the fast joint estimation for time delay and arrival direction is accomplished. The theoretical analysis and simulation results show that the proposed algorithm can significantly reduce the computational complexity and guarantee that the estimation accuracy is not only better than estimating signal parameters via rotational invariance techniques (ESPRIT) algorithm and 2D matrix pencil (MP) algorithm but also close to 2DMUSIC algorithm. Moreover, the proposed algorithm also has certain adaptability to multipath environment and effectively improves the ability of fast acquisition of location parameters.
1. Introduction
Wireless target localization has found widespread application requirements in military and civilian fields. Parameter estimation is an important prerequisite for positioning algorithm, which determines the overall effectiveness of the positioning system. The location parameters include time of arrival (TOA), direction of arrival (DOA), time difference of arrival (TDOA), and received signal strength. The key technique of location finding based on TOA method is to estimate the propagation delay of the radio signal arriving from the direct lineofsight (DLOS) propagation path accurately. The existing TOA methods mainly include crosscorrelation method [1], multiple signal classification (MUSIC) algorithm [2], delay estimation algorithm based on Markov Monte Carlo [3], propagator algorithm [4], and matrix pencil algorithm [5]. If the array antenna can be used to obtain the signal DOA for joint estimation [6, 7], then it will effectively reduce the number of nodes and system overhead of the localization system. In theory, a reference node can lock the location of the target source.
OFDM technology is a multicarrier modulation and highspeed transmission technology, which has been widely used in underwater acoustic communication system, IEEE 802.11a/g/n/ac [8], OFDM new system radar [9], WiMAX system [10], 3GPP LTE/LTEAdvanced [11], 5G mobile communication systems, and so on. However, the demand for target location information of OFDM technology is also increasing, especially in indoor, underground places, and other environments. Therefore, using OFDM array antenna receiving system to achieve accurate positioning of the target signal has become a hot spot of current research.
Due to the limitation of bandwidth, sampling rate, and the attributes of narrowband signals, OFDM signals cannot achieve joint estimation of TOA and DOA employing the triangle geometry relation as in [12, 13]. Ni et al. [14] introduces a TOA estimation algorithm for OFDM wireless signal, which combines coarse estimation with fine estimation. The algorithm needs to use the technologyguided detection to suppress multipath propagation, but the positioning accuracy is limited. Three methods including peak detection, modified maximum peak leak rate detection, and channel frequency domain response reconstruction are proposed in [15]. In the multipath environment of WINNER A1 LOS channel, the channel frequency response reconstruction method can provide the best performance, but this method is mainly based on the received preamble sequence. In the environment of indoor multipath, the twodimensional matrix pencil (2D MP) algorithm is presented in [16] to achieve the joint estimation of time delay and arrival angle, which has low computational complexity, but sacrifices the effective array element and bandwidth. Zhou et al. [17] provide a joint estimation algorithm of delay and DOA, which uses the frequency domain data of single antenna for MUSIC time delay estimation. This algorithm uses the array data for MUSIC DOA estimation, but it fails to give a matching method of time delay and arrival angle, and its computational complexity is high. Ba et al. [18] present a joint estimation of delay and angle using 2MUSIC for twodimensional searching, which solves the pairing problem, but its computational complexity is very high, which makes it hard to apply in practice.
In this paper, a fast estimation method of spacetime twodimensional positioning parameters based on Hadamard product is presented, which can offer the closedform solution of parameter estimation directly without pseudo spectral peak search. This method can implement parameter pairing and improves the efficiency significantly. The computational complexity analysis, CRB derivation, and simulation are also presented. The simulation results show that the proposed algorithm has higher precision of parameter estimation close to the 2DMUSIC algorithm, but the computational complexity is greatly reduced, which effectively improves the ability of fast acquisition of the positioning parameters.
The rest of the paper is organized as follows: Section 2 introduces the data model. Section 3 presents the method of fast estimation and the computational complexity analysis in detail. The CRB for the method is derived in Section 4. Simulation and experimental results are presented in Section 5. Conclusions are drawn in Section 6.
2. Data Model
Considering a node equipped with an equidistant line array composed of antenna elements, the distance between the signal source and the antenna array is far enough, and and represent DOA and propagation delay of the multipath signal, respectively. is the OFDM symbol transmitted with subcarriers. On the account of the wireless multipath propagation, the received signal at the antenna element can be formulated as where is the number of the multipath components and and are the complex attenuation of the path and propagation delay of the path in antenna, respectively. The phase of the complex attenuation is normally assumed random from one snapshot to another with a uniform probability density function [19]. is additive Gauss white noise. The time delay in each antenna element depends not only on the propagation delay but also on the direction of arrival. In particular, for a uniform linear array, the propagation delay associated to be the arriving path in antenna is given by , with being the distance between adjacent array elements in the array and the speed of light. After the Fourier transform of the received signal , the sampled discrete frequency domain channel response of the subcarrier at the antenna element is given by where . In (2), it is clear that the effect of time delay coming
from the antenna array is negligible compared with , so this formula can be further simplified as where denotes additive Gauss white noise with mean zero and variance at the subcarrier and the antenna element, , is the carrier frequency of OFDM signal, is the net data length of OFDM symbol, and is the sampling interval. We can then write this formula (3) in a vector form. where
Moreover, , set , and denotes Hadamard product. Then,
So, the frequency domain response vector of extended array elements can be expressed as where
By eigenvalue decomposition of , we can obtain mutually orthogonal subspace and noise subspace based on Hadamard productextended signal. Using spacetime pseudo spectral search can estimate and , but this method is very timeconsuming, which costs large amount of calculation, and is hard to practical application. Here, the fast estimation method of twodimensional positioning parameters based on Hadamard product is discussed.
3. Fast Estimation Method of TwoDimensional Parameters
3.1. Algorithm Description
According to (7), the frequency domain estimation vector of the reference antenna element is obtained as where and are the estimation vectors and response vectors of the channel in frequency domain, respectively, is Vandermonde matrix of the frequency points at the reference antenna element, and , , , and are the additive Gauss white noise vector.
For the reference antenna element, the eigendecomposition of the covariance matrix with P snapshot measurement data can be expressed as where and are, respectively, the signal subspace and noise subspace and and are the signal eigenvalue diagonal matrix and noise eigenvalue diagonal matrix. By using the noise eigenvector based on Hadamard product to extract space parameter information, MUSICtype algorithm function can be constructed as
The zero value of the function gives the estimation of the time delay, which is equivalent to the value obtained by pseudo spectral peak search. Let and be the power term of , and (11) is not the polynomials of . Then, , the polynomials of RootMUSIC for TOA, can be expressed as
Then, by utilizing the sampled frequency points at the reference element, the corresponding delay estimation can be obtained as where are roots which are the closest to the unit circle in (12).
Similarly, the DOA estimation of the frequency response of the channel at the first frequency point of the M antenna elements can be obtained by the above algorithm.
Since the delay and DOA are estimated separately, the corresponding parameter pairing is required.
3.2. Parameter Pairing
By using the frequency domain channel responses at the reference array element and the next element, the matrix is constructed as where
The covariance matrix of can be an eigenvalue decomposed as , where and are, respectively, the signal subspace and the noise subspace; obviously, the column vector and noise subspace are orthogonal, which make . In order to achieve the pairing of delay and arrival angle, the cost function can be obtained by
So the pairing of the estimated parameters is to minimize the cost function
Firstly, select a fixed from , and let traverse , then the corresponding to the minimum value of the cost function is the correct pairing value.
3.3. Computational Complexity Analysis
The computational complexity of the algorithm in this paper consists of the following four aspects: the estimation of the covariance matrix and , the computational complexity is and , respectively; the eigenvalue decomposition of the covariance matrix and , the computational complexity is and ; the complexity of constructed polynomial is and ; the computational complexity of parameter pairing is . So the total complexity of the algorithm is .
The total complexity of using 2DMUSIC algorithm is , and and are, respectively, the number of grid of delay and angle of arrival search, that is, the range of measurement search is divided by the search step. The computational complexity of ESPRIT algorithm is . Table 1 compares the computational complexity of the three algorithms.

Assuming , , , , and , the computational complexity curves of the subcarrier number in the three algorithms are plotted. It can be seen from Figure 1 that the computational complexity of this paper is slightly higher than that of the ESPRIT algorithm, but much lower than that of the 2DMUSIC algorithm, which greatly reduces the amount of computation.
4. The CramerRao Bound
The CRB is the lower bound of minimum variance which an unbiased estimation can achieve. The following is to derive the CRB of the model. Let the parameter vector be defined as , where . According to (7) and the relevant initial conditions, the joint probability density function of is
Thus, the loglikelihood function of (19), omitting the constant term, is
To calculate the Fisher information matrix (FIM), the derivatives of (20) with respect to , the real part of , the imaginary part of , and can be obtained as
Since is a twodimensional vector, the derivatives of (20) with respect to are different from the ones in the anglespread only model. and similarly, where is the derivative with respect to of the column of . Similarly, is the derivative with respect to of the column of . Due to , then
Further, written more compactly where and . Let and , then we can obtain
By using the results proven in [20], the following expressions can be attained
Finally, the FIM for the parameters is obtained. Using the results in [20], the CRB of the parameters can be shown as which completes the proof.
5. Simulation Results
In this section, the performance of fast estimation of twodimensional positioning parameters of TOA and DOA is studied. Monte Carlo simulation is utilized to verify the practicability and robustness of the proposed algorithm. The relevant parameters of narrowband array antenna in OFDM system are set as follows: the system bandwidth is , the FFT cycle is , the number of subcarriers is , the carrier frequency is , the number of elements is , the array spacing is meters, and so on.
5.1. Simulation 1
In the case of low SNR at 0 dB and 5 dB, assuming that the multipath number of the received signals is , the arrival time is , , and , respectively; the direction of arrival is , , and , respectively. Simulation time of the algorithm is , the scatter diagrams of the joint estimation of TOA and DOA combined with three multipath components are obtained, Figures 2 and 3 show that the algorithm has a good performance for the joint estimation at low SNR conditions.
5.2. Simulation 2
Under different SNRs, other conditions are the same as simulation 1, and the scatter plots of TOA and DOA can be obtained. It can be seen from Figure 4 and Figure 5 that with the increase of the SNR, the dispersion of TOA and DOA gradually tends to a point, and the estimation precision is higher. At low SNR, the estimation accuracy amplitudes of TOA and DOA are all within a unit.
5.3. Simulation 3
At low SNR of SNR = 5 dB, other conditions are the same as simulation 1, and the estimation error of TOA and DOA is defined as and , respectively. Then, the scatter diagram of , , and corresponding to three multipath components is shown in Figure 6, and the corresponding , , and distribution are shown in Figure 7. It can be seen from Figures 6 and 7 that the algorithm at low SNR can almost achieve the unbiased estimation of TOA and DOA.
5.4. Simulation 4
Define the root mean square error (RMSE) as , where is the parameter estimated value obtained by the simulation and is the true value of the corresponding parameter. The TOA and DOA estimation performances of the respective first path are shown in Figures 8 and 9, respectively. The number of multipath is , , and , respectively. The results are based on 100 Monte Carlo simulations. As expected, the TOA and DOA estimation performances of the first path decrease with the increase of the number of multipath. At low SNR of SNR = 5 dB, the estimation accuracy under multipath conditions can still meet the positioning requirements, so the algorithm has certain adaptability to multipath environment.
5.5. Simulation 5
Under different SNRs, other conditions are the same as simulation 1. The proposed algorithm is compared with 2DMUSIC algorithm, ESPRIT algorithm, 2D MP algorithm [16], and 1D MP algorithm [13]. The TOA and DOA RMSE curves of each algorithm’s first path are shown in Figures 10 and 11, respectively, and compared with the CRB of this model. As shown in Figures 10 and 11, the performance of the proposed algorithm is obviously superior to that of ESPRIT algorithm and 2D MP algorithm, and very close to 2DMUSIC algorithm, and the accuracy of parameter estimation completely meets the positioning requirements.
Since the 1D MP algorithm is based on a single snapshot data, little information is available, and the 1D MP algorithm is the worst in low SNRs. However, the 2D MP algorithm extends the information, and the availability of the information is relatively increased so that the performance of 2D MP algorithm is better than that of 1D MP algorithm. In essence, the proposed algorithm is another form of MUSIC algorithm and has a similar estimation performance as 2DMUSIC algorithm, of which the complex twodimensional pseudo spectral peak search is avoided in the proposed algorithm. Therefore, it can realize the fast estimation of twodimensional location parameters.
6. Conclusion
The fast estimation of location parameter is a necessary condition for the quality of wireless positioning system. As the fact that 2DMUSIC algorithm needs a long time twodimensional pseudo spectral peak search, which makes it difficult to be applied in practice, a fast estimation method of twodimensional positioning based on Hadamard product in OFDM system is proposed. The construction of the model, the theoretical analysis, and the process of CRB derivation are presented. The computation complexity of the proposed method by solving the closedform solution and parameter pairing is much lower than that of 2DMUSIC algorithm. The simulation results show that the parameter estimation performance of the proposed algorithm is not only better than that of ESPRIT algorithm and 2D matrix pencil algorithm, but also close to that of 2DMUSIC algorithm, and it has certain adaptability to the multipath environment and certain practical value.
Notation
:  Matrix transpose 
:  Matrix conjugate 
:  Matrix conjugate transpose 
:  Statistical expectation 
:  Hadamard product 
:  Kronecker product. 
Conflicts of Interest
The authors declare that they have no conflicts of interest.
Acknowledgments
This work is supported in part by the National Natural Science Foundation of China (Grant no. 61401513) and the National Science and Technology Major Project of the Ministry of Science and Technology of China (Grant no. 2013ZX03006003006).
References
 A. Makki, A. Siddig, M. Saad, J. R. Cavallaro, and C. J. Bleakley, “Indoor localization using 802.11 time differences of arrival,” IEEE Transactions on Instrumentation and Measurement, vol. 65, no. 3, pp. 614–623, 2016. View at: Publisher Site  Google Scholar
 F.X. Ge, D. Shen, Y. Peng, and V. O. K. Li, “Superresolution time delay estimation in multipath environments,” IEEE Transactions on Circuits and Systems I: Regular Papers, vol. 54, no. 9, pp. 1977–1986, 2007. View at: Publisher Site  Google Scholar
 L. Jing, Z. YongJun, and L. DongHai, “Time delay estimation using Markov chain Monte Carlo method,” Acta Physica Sinica, vol. 63, no. 13, article 130701, 2014. View at: Publisher Site  Google Scholar
 F.q. Wang and X.f. Zhang, “Improved propagator methodbased joint TOA and DOA estimation in impulse radio ultra wide band,” Journal of Electronics & Information Technology, vol. 35, no. 12, pp. 2954–2959, 2013. View at: Publisher Site  Google Scholar
 A. Gaber and A. Omar, “Utilization of multipleantenna multicarrier systems and NLOS mitigation for accurate wireless indoor positioning,” IEEE Transactions on Wireless Communications, vol. 15, no. 10, pp. 6570–6584, 2016. View at: Publisher Site  Google Scholar
 A.J. van der Veen, M. C. Vanderveen, and A. Paulraj, “Joint angle and delay estimation using shiftinvariance techniques,” IEEE Transactions on Signal Processing, vol. 46, no. 2, pp. 405–418, 1998. View at: Publisher Site  Google Scholar
 R. Shafin, L. Liu, Y. Li, A. Wang, and J. Zhang, “Angle and delay estimation for 3D massive MIMO/FDMIMO systems based on parametric channel modeling,” IEEE Transactions on Wireless Communications, vol. 16, no. 8, pp. 5370–5383, 2017. View at: Publisher Site  Google Scholar
 “Part 11: wireless LAN medium access control (MAC) and physical layer (PHY) specifications, amendment 4: enhancements for very high throughput for operation in bands below 6 GHz,” in IEEE 802.11ac Std, 2013. View at: Google Scholar
 K. Huo and J. Zhao, “The development and prospect of the new OFDM radar,” Journal of Electronics & Information Technology, vol. 37, no. 11, pp. 2776–2789, 2015. View at: Google Scholar
 “Advanced air interface, mobile WiMax (release 2),” in IEEE Standard 802.16m, 2011. View at: Google Scholar
 “Study on Channel Model for Frequency Spectrum Above 6 GHz (Release 14),” Tech. Rep., document TR 38.900, 3GPP, 2016. View at: Google Scholar
 F. Q. Wang, X. F. Zhang, and F. Wang, “RootMUSICbased joint TOA and DOA estimation in IRUWB,” Journal on Communications, vol. 35, no. 2, pp. 137–145, 2014. View at: Google Scholar
 R. Ding, Z. H. Qian, and X. Wang, “UWB positioning system based on joint TOA and DOA estimation,” Journal of Electronics & Information Technology, vol. 32, no. 2, pp. 313–317, 2010. View at: Publisher Site  Google Scholar
 H. Ni, G. Ren, and Y. Chang, “A TDOA location scheme in OFDM based WMANs,” IEEE Transactions on Consumer Electronics, vol. 54, no. 3, pp. 1017–1021, 2008. View at: Publisher Site  Google Scholar
 Z. He, Y. Ma, and R. Tafazolli, “Improved high resolution TOA estimation for OFDMWLAN based indoor ranging,” IEEE Wireless Communications Letters, vol. 2, no. 2, pp. 163–166, 2013. View at: Publisher Site  Google Scholar
 A. Gaber and A. Omar, “A study of wireless indoor positioning based on joint TDOA and DOA estimation using 2D matrix pencil algorithms and IEEE 802.11ac,” IEEE Transactions on Wireless Communications, vol. 14, no. 5, pp. 2440–2454, 2015. View at: Publisher Site  Google Scholar
 B. Zhou, C. Jing, and Y. Kim, “Joint TOA/AOA positioning scheme with IPOFDM systems,” Wireless Personal Communications, vol. 75, no. 1, pp. 261–271, 2014. View at: Publisher Site  Google Scholar
 B. Ba, G. Liu, T. Li, Y. Lin, and Y. Wang, “Joint for time of arrival and direction of arrival estimation algorithm based on the subspace of extended hadamard product,” Acta Physica Sinica, vol. 64, no. 7, pp. 0784031–0784039, 2015. View at: Google Scholar
 X. Li and K. Pahlavan, “Superresolution TOA estimation with diversity for indoor geolocation,” IEEE Transactions on Wireless Communications, vol. 3, no. 1, pp. 224–234, 2004. View at: Publisher Site  Google Scholar
 P. Stoica and A. Nehorai, “Music, maximum likelihood, and CramerRao bound,” IEEE Transactions on Acoustics, Speech, and Signal Processing, vol. 37, no. 5, pp. 720–741, 1989. View at: Publisher Site  Google Scholar
Copyright
Copyright © 2018 Haiwen Li 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.