Signal Processing - March 2017 - 45

for the first experiment, the frequency separation in the angular dimension is the largest, whereas for the second experiment
the frequency separation in the range dimension is the largest. For all dimensions, the separations are below the resolution limit except for x-positions below 18 m, where the angular
separation exceeds the resolution limit for the first experiment.
Therefore, we expect that conventional radar processing will
fail for both experiments, except for x-positions below 18 m in
the first experiment. For the first experiment, we further expect
that high-resolution processing in the spatial domain will perform well due to the dominating angular separation and that
high-resolution processing in the Fourier domain selects the
angular dimension as resolution dimension, leading to a similar performance. For the second experiment, we expect that
high-resolution processing in the Fourier domain correctly
selects the range dimension as resolution dimension and will
thus provide resolved and accurate frequency estimates. In contrast, we expect that high-resolution processing in the spatial
domain performs significantly worse due to the predefined
resolution dimension and the much smaller angular separation.
Figure 4 shows the resulting position estimates in Cartesian coordinates for the first experiment, which are obtained
with (a) conventional radar processing, (b) high-resolution frequency estimation in the spatial domain, and (c) high-resolution
frequency estimation in the Fourier domain with optimal
selection of resolution dimension. The true target positions are
shown as dashed vertical lines. As expected, conventional processing fails to resolve the two targets except for x-positions
below 18 m. In contrast, both high-resolution approaches are
able to resolve the targets and provide accurate position estimates. Note that this means that the resolution dimension was
correctly selected by the approach proposed in the section "Optimal Selection of the Resolution Dimension."
Figure 5 shows the corresponding results for the second experiment. Here, conventional processing fails for all x-positions. The
high-resolution approach in the spatial domain breaks down for
x-positions above 30 m due to the small frequency separation
in the angular dimension. Only the high-resolution, Fourierdomain approach can resolve the two targets for all x-positions,
as it exploits the much larger range separation. Note that this is
enabled by the approach proposed in the section "Optimal Selection of the Resolution Dimension," which correctly selects the
range dimension as the resolution dimension.

Conclusions
We have considered automotive radar as a key technology for
ADASs and HAD. Current limitations with respect to the
requirements of the automotive industry can be met by using
high-resolution frequency estimation. We have described conventional automotive radar processing and pointed out use
cases in which it is bound to fail. We have presented a flexible
framework for computationally efficient high-resolution
frequency estimation as an enhancement to conventional radar
processing. Real data from a series-production automotive
radar sensor have been presented to show the effectiveness of
the presented approach.

Authors
Florian Engels (florian.engels@continental-corporation.com)
received his M.Sc. and Dr.-Ing. degrees, both in electrical engineering, from Karlsruhe Institute of Technology [formerly
Universität Karlsruhe (TH)], Germany, in 2008 and 2016,
respectively. Since 2008, he has been with Continental, Adaptive
Distance Control Systems GmbH, Lindau, Germany, as a radar
algorithm development engineer and, since 2015, as a leading
expert for radar signal processing. His research interests include
multidimensional frequency estimation with applications in
automotive radar.
Philipp Heidenreich (philipp.heidenreich@opel.com)
received his Dipl.-Ing. and Dr.-Ing. degrees, both in electrical
engineering, from Technische Universität Darmstadt, Germany,
in 2006 and 2012, respectively. From 2007 to 2012, he was with
the Signal Processing Group, Technische Universität Darmstadt,
as a research scientist. From 2012 to 2015, he was with
Continental, Adaptive Distance Control Systems GmbH,
Lindau, Germany, as a radar algorithm development engineer.
Since 2015, he has been with General Motors Europe, Adam
Opel AG, Rüsselsheim, Germany, as a systems engineer for
automotive active safety applications. His research interests
include statistical and array signal processing with applications
to automotive radar. He received the Best Student Paper Award
at the Fifth Biennial Workshop on Digital Signal Processing for
In-Vehicle Systems in 2011, in Kiel, Germany. He is a Member
of the IEEE.
Abdelhak M. Zoubir (zoubir@spg.tu-darmstadt.de)
received his Dr.-Ing. degree from Ruhr-Universität Bochum,
Germany, in 1992. He was with Queensland University of
Technology, Australia, from 1992 to 1998, where he was an
associate professor. In 1999, he joined Curtin University of
Technology, Australia, as a professor of telecommunications. In
2003, he moved to Technische Universität Darmstadt, Germany,
as a professor of signal processing and the head of the Signal
Processing Group. His research interest lies in statistical methods for signal processing with emphasis on bootstrap techniques, robust detection and estimation, and array processing
applied to telecommunications, radar, sonar, automotive monitoring and safety, and biomedicine. He has served on publication boards of various publications, most notably as the
editor-in-chief of IEEE Signal Processing Magazine (2012-
2014). He is a Fellow of the IEEE and is an IEEE Distinguished
Lecturer (class of 2010-2011).
Friedrich K. Jondral (Friedrich.Jondral@kit.edu) received
the diploma in mathematics and the doctoral degree in natural
sciences from the Technische Universität Braunschweig,
Germany, in 1975 and 1979, respectively. Since 1993, he has
been a full professor and the director of the Communications
Engineering Laboratory at the Karlsruhe Institute of
Technology [formerly Universität Karlsruhe (TH)], Germany,
where he served as dean of the Department of Electrical
Engineering and Information Technology from 2000 to 2002.
His current research interests include software-defined and cognitive radio, cooperation and coordination, signal analysis, pattern recognition, network capacity optimization, and dynamic

IEEE SIgnal ProcESSIng MagazInE

March 2017

http://www.M.Sc

Table of Contents for the Digital Edition of Signal Processing - March 2017