Upload
xinying
View
216
Download
4
Embed Size (px)
Citation preview
A DOA Estimation Algorithm of Virtual Array Based on Beam forming
Qingwei Li College of Information and
Communication Engineering Harbin Engineering University
Harbin, China [email protected]
Yi Jiang College of Information and
Communication Engineering Harbin Engineering University
Harbin, China [email protected]
Xinying Diao College of Information and
Communication Engineering Harbin Engineering University
Harbin, China [email protected]
Abstract—Virtual Array Transformation of DOA Algorithm is widely used in actual engineering DOA estimation of the uniform linear array (ULA) of spacing of array elements greater than half a wavelength or any non-uniform linear array (NULA).However, when virtual array is transformed, if there is a signal outside the current transformation sensor, a serious impact will be had on virtual array DOA estimation performance of the signal in the current transformation interval. In this paper, a virtual array transformation algorithm is proposed based on beam forming methods, which can effectively solve the existing problems of traditional virtual array transformation algorithms, and the mathematical model and test simulation results of the algorithm are given.
Keywords-DOA estimation; virtual array transformation; transformation sectors; beam forming
I. INTRODUCTION Conventional MUSIC algorithm and spatial smoothing
algorithm are always considered for ULA of Array element spacing 0.5d λ≤ . But it is very hard to ensure 0.5d λ≤ ULA in actual engineering, which makes the classic MUSIC algorithm and spatial smoothing DOA estimation algorithm restricted. However, the virtual array transformation method can transform any NULA or ULA with spacing of array elements 0.5d λ> to ULA of spacing of array elements
0.5d λ≤ , what not only makes the study of previously ULA DOA estimation algorithm can be used, but also increase the aperture of array for using the virtual array elements.
When virtual transformation is taken, we should calculate the transformation matrix. If the transformation interval is too large, the transformation matrix obtained makes the error of DOA estimated angle too large. Therefore, the intervals divided of the scanned area should be small enough to ensure the accuracy of DOA estimation of virtual transform in the divided interval. When a scan range is selected, first we divide the range into several intervals. When we take virtual transform in a certain interval, if a signal exists outside this interval, the DOA estimation of virtual transformation of signals in this interval will be inaccurate.
As a result, this paper presents a virtual array DOA method based on beam-forming. Before we have the DOA estimation of virtual transform in a certain interval, we can first use the
weighting factors for beam-forming to weight every sub-array. Then we can filter the signals outside the current transform interval and can have accurate DOA estimation of virtual transformation.
II. THE PRINCIPLE OF VIRTUAL ELEMENTS TRANSFORMATION
In this section we describe the principle of virtual array transformation. Let us divide the field of view of the array into u sectors, which is 1 2 u, , ........., .Now we discuss the principle of virtual transformation in the j th sector j . And then the sector is
[ , , 2 ,...... ]l l l rθ θ θ θ θ θ= + Δ + Δj (1)
where θΔ is the step of angle. Then the steering vector of actual array can be obtained
[ ( ) ( ) ......... ( )]L L rθ θ θ θ= + ΔA (2)
And the steering vector of virtual array is
[ ( ) ( ) ......... ( )]L L rθ θ θ θ= + Δ (3)
In other words, A is the response of the real array to signals in the sector j ,and is the response of interpolated array to the same signals. The spacing of the virtual array elements is a half wavelength. Then, there exists a constant matrix B between the actual array and the virtual array, which can be written as
= ⇔ =H HBA A BAA AA (4)
From above formula, we can obtain that
H H -1B = AA (AA ) (5)
Assume that the actual data covariance matrix is R and the noise of system is additive Gaussian white noise, the average power of which is 2δ . Then
2δ= +SR AR A I (6)
Hence the data covariance matrix of virtual array is
945978-1-4244-8165-1/11/$26.00 ©2011 IEEE
2δ+H H HSR = BRB = BAR A B BB (7)
Because ≠HBB I , and from (6) and (7), we know that the white noise has become color noise. Therefore the noise received is independent. So we should revise the matrix of transformation B as that
H -1/ 2T = (BB ) B (8)
We can prove that
HTT = I (9)
So, the covariance matrix of rectification is
2δ+H H HSR = TRT = TAR A T I (10)
III. DOA ESTIMATION OF VIRTUAL ARRAY TRANSFORMATION BASED ON BEAM FORMING
2x 3x1x Mx
1v 2v 3v 4v pv
m u×
Figure 1. The structure chart of DOA estimation of virtual array
transformation based on beam forming
Fig. 1 shows a Uniform Linear Array of M elements, the spacing of which is d .Let us divide the array into numbers of sub-arrays as follow:
We give a group of serial number to the elements of the array from the left one to the right. So we can get the serial number is 1 2 …….M. The length of every sub-array ism . Hence the serial number of the k th sub-array can describe as , 1,......, 1k k k m+ + − . Assuming we can divide the array into p sub-arrays, therefore we can obtain that
1p M m= + − (11)
According the description of above, we have divided the field of view into u sectors, which is 1 2 u, , ........., .
Assume there are N signals received by the elements of the array. The angles of these signals are 1 2( , ,....... )Nθ θ θ .And
the j th signal ( )js t is located in the j th sector j . The other incoming signals are located in others sectors.
The general formula of incoming signals can be written as
( ( ))( ) ( ) , 1, 2.........ic ij w t ti is t u t e i Nφ+= = (12)
According to the Snapshot representation of Array signal processing we can obtain that
;= +X AS N (13)
where [ ]1 2 ......... TNs s s=S and
1 2[ ( ) ( ) ....... ( )]Nθ θ θ=A , which can be called Array Flow Pattern Matrix. And ( )iθ can be called Steering Vector. Where 1,2.........i N= .
In fig.1, we use kx to express the data received by the k th element. And kX stands for the data vector received by the k th sub-array. Then
( 1) ;k− +kX = AD S N (14)
where 1,2....k p= and [ ]1 1...... Tk k k k mx x x+ + −=X ,
1
2
0 00 0
0 0 N
j
j
jN N
e
e
e
β
β
β
−
−
−×
=D ,
and2 sin
, 1,2......ii
di N
π θβλ
= = .
For the u sectors, there are u groups of weighting factors corresponding to them. The weighting factors corresponding to every sector is calculated by beam forming algorithm. And the beam pointing of every group weighting factor is to corresponding sector.
In the net of beam weighting,
[ ]= 1 2 uT W W ....... W (15)
where iW is the weighting factor corresponding to the i th
sector. And ,1 ,2 ,...... , 1, 2......,T
i i i mw w w i u= =iW .So the combining data of the k th sub-array is
kv = Ti KW X (16)
where kv is the bombing data of the k th sub-array, and
[ ]1 1...... Tk k k k mx x x+ + −=X
We can work out
,1 ,2 1 , 1.........k j k j k j m k mv w x w x w x+ + −= + + + (17)
where 1,2.....k p= .
946
If we write the data vector of each sub array as 1 2 .......
T
pv v v=V , we can obtain the actual data covariance matrix of all the sub arrays R
[ ]HE=R V * V (18)
And then, according to (5) and (8) we can obtain the matrix of transformation. After that we can work out the virtual data covariance matrix R by (10). In this procedure we expand p branches of date to 2 1p − . Next, we can get the eigenvalues
of R and then obtain the angle of the incoming signal ( )js t of current transformation sector j by MUSIC algorithm.
By the same method and step, we can get the angles of different incoming signals in different sectors, avoiding the impact to the signal in current transformation sector by signals in other sectors.
IV. SIMULATION
Assuming that the scanning range is [ 90 , 90 ]° °− − in the test, the entire scanning range is evenly divided into three sectors that are [ 90 , 30 ]° °= − −1 , [ 30 ,30 ]° °= −2
[30 ,90 ]° °=3 .
The same simulation conditions of all the experiments are given as follows
The spacing of the array elements is ULA of spacing of elements 0.7d λ= . The number of array elements is
12M = .And the number of every sub array elements is 6m = . The number of sub arrays is 1 7p M m= + − = .The SNR is 0dB, step of angle is 0.01°and sampling snapshot is 500. The incoming signals are two and their angles are 0°and 80°.
4.1Take the transformation sector as [ 30 30 ]° °= −2 . We directly have virtual transformation in the sector without any preprocessing. The result is given as follows.
-30 -20 -10 0 10 20 30-15
-10
-5
0
5
10
15
20DOA estimation of Virtual transformation without any preprocessing
Angle(°)
Spa
tial S
pect
rum
(dB
)
Figure 2. The figure of DOA estimation of virtual transformation without
any preprocessing
From the result of simulation we can have a conclusion. For there is a signal outside of the current transformation sector, as a result, there is a pseudo peak of wave. If we take virtual transformation without any processing to the signal outside the current sector, the algorithm of virtual transformation of DOA estimation will lose its efficacy.
.2 Validation of interference suppression in DOA estimate by beam forming
All the sub arrays are Hamming Weighted, beam pointing to 0 . Results show as follow:
-100 -80 -60 -40 -20 0 20 40 60 80 100-10
-5
0
5
10
15
20
25DOA estimation after beam forming
Angle(°)
Spa
tial S
pect
rum
(dB
)
Figure 3. DOA estimatation after beam forming
Fig.3 shows the signal outside the current selected sector 2 can be filtered efficiently by weighting factors to every sub
array for beam forming.
4.3 Before taking virtual transformation in the simulation condition of 4.1, first we give Hamming Weighting factors to every sub array for beam pointing to 0 and take [ 30 30 ]° °= −2 as the transforming sector. The result shows as follows.
-30 -20 -10 0 10 20 30-10
-5
0
5
10
15
20
25
30DOA estimation based on virtual array of beam forming
Angle(°)
Spa
tial S
pect
rum
(dB
)
Figure 4. DOA estimatation based on virtual array based on beam forming
947
Compared with Fig. 2, Fig. 4 shows interference of signals outside the current transformation sector can be effectively restrained by filtering in spatial domain with beam forming for each sub array before transformation.
Fig. 4 also shows that when SNR=0 dB, high precision of DOA estimation can be achieved by beam forming of sub array which has strong filtering capability for interference. As to the signal in the sector [30 ,90 ]° °=3 , the same method can be applied into use. Select the transform sector [30 ,90 ]° °=3 , sub arrays are weighted by the weighting factors of beam pointing to 60 . And the result shows as follow:
30 40 50 60 70 80 90-10
-5
0
5
10
15
20
25
30DOA estimatation based on virtual array of beam forming for signal of 80°
Angle(°)
Spa
tial S
pect
rum
(dB
)
Figure 5. DOA estimatation based on virtual array of beam forming for
signal of 80
V. CONCLUSIONS A new method that applies beam forming to DOA
estimation of virtual array is proposed in this paper. At the time, this paper gives out the mathematical model of the virtual array of DOA estimation based on beam forming and the results of simulation .By weighting every sub array with corresponding weighting factors, it can effectively filter the signal outside the current transforming sector. Not only can it effectively overcome the shortage of directly taking virtual transformation, but also can it filter all the noise outside the beam pointing. For the whole array is divided into different sub-arrays, which makes array aperture reduced. But this can be compensated through next virtual array transformation.
ACKNOWLEDGMENT This paper is funded by the International Exchange
Program of Harbin Engineering University for Innovation-oriented Talents Cultivation.
REFERENCES [1] Joseph C.Liberti,and Theodore S.Rappaport, Smart Antennas for
Wireless Communications:IS-95 and Third Generation CDMA Applications. Beijing: China Machine Press, August,2002.
[2] WANG Yongling, Chen Hui, and Wan Qun, Theory and Algorithm of Spatial Spectrum Estimation. Beijing:tsinghua university press, November, 2004
[3] DIAO Ming, and WANG Yanwen, “Direction of arrival estimation of coherent sources based on arbitrary plane arrays”,Journal of Marine Science and Application, vol. 4,September, pp.53-57,2005.
[4] Wang Yongliang,Chen Hui, and Wan Shanhu, “An effective DOA method via virtual array transformation”,SCIENCE IN CHINA, vol. 44, pp.75-82,February,2001.
[5] ZHOU Zou, WANG Hong-yuan,and GUO Yue, “Improved Super-Resolution MUSIC Algorithm with Spurious Peaks Restraining Based on Virtual Arrays”, Microelectronics & Computer, vol. 24, pp.5-7,2007.
[6] Ding Pu, “Research of space domain virtual array DOA algorithms in smart antenna system”, Information & Comunications, vol. 6 pp.28-30,2007.
948