IEEE - Aerospace and Electronic Systems - November 2019 - 55

Huizing et al.
integration time in the Doppler filtering had a higher
impact on the classification performance than the type of
the deep neural network (CNN versus LSTM-RNN) in the
classifier. The use of simulated spectrograms to classify
measured spectrograms requires a higher fidelity modeling
of the micro-Doppler signals than the modeling approach
based on principal scatterers used in this paper.
When a cognitive radar measures a micro-Doppler
spectrogram from a target class that is not represented in
the training set, an anomaly detection algorithm could provide a trigger to the cognitive radar scheduler to collect
micro-Doppler spectrograms from this unknown target.
The performance of two deep learning techniques, Softmax and GANomaly, for the detection of spectrograms
from unknown target classes has been compared. Both
techniques had a similar performance with respect to the
probability of detecting unknown classes with the GANomaly technique showing a larger performance variation
due to the difficulty in training adversarial networks. Different GAN architectures for anomaly detection that are
easier to train, such as a Wasserstein GAN may provide a
better performance [22].
A denoising adversarial auto-encoder has been trained
to remove noise from spectrograms. An improvement of
20 dB in the SNR of a measured spectrogram has been
observed. If this denoising technique does not alter the
micro-Doppler characteristics significantly, a considerable
increase in the range at which a cognitive radar can recognize mini-UAVs could be achieved.
To mitigate the effects of class imbalances in the number of training examples on the performance of a classifier, an InfoGAN can be used to generate additional
spectrograms of underrepresented classes. A quantitative
study concerning the benefits of this approach will be conducted in the future research.
The overall conclusion of this paper is that deep learning techniques have a high potential for improving the
classification of mini-UAVs using micro-Doppler spectrograms in cognitive radar. However, the selection and tuning of the appropriate neural network architectures for
different processing tasks, such as classification and
denoising is tedious and time consuming. A better understanding of the theory behind deep neural networks, potentially enabled by the link with compressive sensing and
sparse signal representations, would not only accelerate
the development process, but also improve the acceptance
of these advanced techniques in military radar
applications.

Dutch Radar Center of Expertise (D-RACE), a strategic
alliance between Thales Nederland B.V. and TNO.

REFERENCES
[1] S. Haykin, Y. Xue, and P. Setoodeh, "Cognitive radar: Step
toward bridging the gap between neuroscience and engineering," Proc. IEEE, vol. 100, no. 11, pp. 3102-3130,
Nov. 2012.
[2] J. R. Guerci, Cognitive Radar: The Knowledge-Aided
Fully Adaptive Approach, Norwood, MA, USA: Artech
House, 2010.
[3] M. S. Greco, F. Gini, P. Stinco, and K. Bell, "Cognitive
radars: On the road to reality: Progress thus far and possibilities for the future," IEEE Signal Process. Mag., vol. 35,
no. 4, pp. 112-125, Jul. 2018.
[4] Y. LeCun, Y. Bengio, and G. Hinton, "Deep learning,"
Nature, vol. 521, no. 7553, pp. 436-444, 2015.
[5] D. A. E. Morgan, "Deep convolutional neural networks for
ATR from SAR imagery," Proc. SPIE, vol. 9475, 2015,
Art. no. 94750F.
[6] Y. Kim and T. Moon, "Human detection and activity classification based on micro-Doppler signatures using deep
convolutional neural networks," IEEE Geosci. Remote
Sens. Lett., vol. 13, no. 1, pp. 8-12, Jan. 2016.
€
[7] M. S. Seyfioglu, S. Z. G€urb€uz, A. M. Ozbayo
glu, and
M. Y€uksel, "Deep learning of micro-Doppler features for
aided and unaided gait recognition," in Proc. IEEE Radar
Conf., 2017, pp. 1125-1130.
[8] R. P. Trommel, R. I. A. Harmanny, L. Cifola, and
J. N. Driessen, "Multi-target human gait classification
using deep convolutional neural networks on microDoppler spectrograms," in Proc. Eur. Radar Conf., 2016,
pp. 81-84.
[9] B. Dekker, S. Jacobs, A. S. Kossen, M. C. Kruithof,
A. G. Huizing, and M. Geurts, "Gesture recognition with a
low power FMCW radar and a deep convolutional neural
network," in Proc. Eur. Radar Conf., 2017, pp. 163-166.
[10] P. Molchanov, K. Egiazarian, J. Astola, R. I. A. Harmanny,
and J. J. M. de Wit, "Classification of small UAVs and birds
by micro-Doppler signatures," in Proc. Eur. Radar Conf.,
2013, pp. 172-175.
[11] R. Harmanny, J. J. M. de Wit, and G. Premel-Cabic,
"Radar micro-Doppler mini-UAV classification using
spectrograms and cepstrograms," Int. J. Microw. Wireless
Technol., vol. 7, no. Special Issue 3/4, pp. 469-477, 2015.
[12] V. C. Chen, F. Li, S.-S. Ho, and H. Wechsler, "MicroDoppler effect in radar: Phenomenon, model, and simulation study," IEEE Trans. Aerosp. Electron. Syst., vol. 42,
no. 1, pp. 2-21, Jan. 2006.

ACKNOWLEDGMENT
TNO has received funds for this study from the Netherlands
Ministry of Defense, under Grant V1512 and Grant V1908.
The study has been conducted within the scope of the

NOVEMBER 2019

[13] P. J. Speirs, A. Schr€oder, M. Renker, P. Wellig, and
A. Murk, "Comparisons between simulated and measured

IEEE A&E SYSTEMS MAGAZINE

X-band signatures of quad-, hexa- and octocopters," in
Proc. 15th Eur. Radar Conf., 2018, pp. 325-328.

55
IEEE - Aerospace and Electronic Systems - November 2019

Table of Contents for the Digital Edition of IEEE - Aerospace and Electronic Systems - November 2019
