UWB Virtual Beamforming Using
Hermite Fractional Delay Filter Sets Combining
Autoregressive Modeling
Qiang Du, Yaoliang Song, Zeeshan Ahmad, Chenhe Ji
School of Electronic and Optical Engineering, Nanjing University of Science and Technology
Nanjing, China
qdu1023@gmail.com, ylsong@mail.njust.edu.cn, engr.zeeshan@hotmail.com, princekevin1984@163.com
Abstract—The digital beamforming (DBF) of ultra-wideband
(UWB) radar is a challenging issue and its performance is closely
affected by the delay compensation filters and the aperture of the
antenna array. In this paper, a high-resolution DBF method for
UWB signals utilizing Hermite fractional delay filter sets
combining a virtual array created by a vector extrapolation
algorithm on autoregressive(AR) models is presented.
Theoretical analysis indicates that the magnitude response and
the group delay of the Hermite fractional delay filter sets are
much more accurate than that of the traditional method such as
Lagrange fractional delay filter over the whole frequency band,
and the adopted virtual array creation algorithm can extend the
aperture of the original array to 6 times. Simulation results of the
DBF performance using UWB linear frequency modulated (LFM)
signals for two different scenarios also demonstrate the
superiority of the proposed method.
Keywords—UWB radar; virtual beamforming; vector
extrapolation; Hermite fractional delay; autoregressive (AR)
modeling
I. INTRODUCTION
UWB radars have been widely used in this decade, but the
DBF for UWB signals is a challenging issue [1]. DBF based
on phase-shift method, which is the most common way for
narrowband radars, will cause pointing deviation when applied
to UWB radars. Thus DBF for UWB signals becomes a
bottleneck problem for UWB radars.
There are two major approaches to implement DBF for
UWB signals: sub-band decomposition in the frequency
domain [2] and time delay compensation in the time domain
[3]. Time delay compensation can be implemented by analog
delay lines [4] or digital interpolation methods [5]. The analog
delay lines method has lower accuracy, higher cost and is
unstable, especially when optical true time delay (OTTD) is
used. The digital interpolation methods, such as fractional
delay filter methods [5-6], can achieve arbitrary delay and can
be implemented much more conveniently and slash the cost,
but the magnitude responses and group delay responses of the
proposed filters in these references are undesirable, when
signal frequencies approach the Nyquist rate[7-8]. The
Hermite interpolation method not only can overcome the
undesirability, but its implementation is also simple as well[9].
The aperture of an array is another important factor
affecting the DBF performance. However, increasing the true
antennas is not practical in many applications. So the concept
of virtual array has been proposed to extend the number of
antennas, consequently increase the aperture of the array[10].
One method for creating virtual measurement data set is to use
an AR model to extrapolate the original one, but this method
only extrapolate all the elements of original measurement data
set one by one [10]. So the vector extrapolation algorithm is
more efficient [11-12].
Hence, we present a new virtual DBF(VDBF) method
based on high-order Hermite fractional delay filter sets
combining a virtual array created by a special vector
extrapolation algorithm on AR models. First the block AR
parameters are estimated from the original measurement data
sets. Then the virtual measurement data sets are created by a
vector extrapolation algorithm. Finally a DBF based on high-
order Hermite fractional delay filter sets is applied to the
virtual measurement data sets. Simulations show that the
proposed method not only improves the spatial resolution of
an ULA significantly, but also eliminates false peaks when
Lagrange Fractional delay filters are applied to a true array.
The low-level signal content is preserved in the virtual
measurement data sets where one or two obscured low-SNR
closely spaced UWB sources are masked by another nearby
dominant source without increasing the number of antennas.
The rest of this paper is organized as follows. Section II
develops a signal model, whereas Section III presents the
proposed VDBF algorithm. To illustrate the high-resolution
capability of resolving low-SNR, closely spaced sources of the
proposed algorithm, computer simulations are conducted, and
the results are given in Section IV. Finally section V offers
some conclusions drawn on the basis of simulation results.
II. SIGNAL MODELING
Consider M sources impinging on a ULA comprised of L
antennas equally spaced by d, where M < L is assumed. The
M signals are coherent, in the far field, arriving from angles
off from normal to the array, and
.
The sampled data of the l-th antenna received signal is
given by
, where
is received analog
signal in Additive Gaussian White Noise (AGWN) before
sampling on this antenna and
. n is the
This work was supported in part by the Natural Science Foundation of
China (Grant No. 61071145 and No. 61271331).