Moving Target Classification in Automotive Radar Systems Using Convolutional Recurrent Neural Networks

Moving target classification is a key ingredient to avoid accident in autonomous driving systems. Recently, fast chirp frequency modulated continuous wave (FMCW) radar has been popularly used to recognize moving targets due to its ability to discriminate moving objects and stationary clutter. In order to protect vulnerable road users such as pedestrians and cyclists, it is essential to identify road users in a very short period of time. In this paper, we propose a deep neural network that consists of convolutional recurrent units for target classification in automotive radar system. In our experiment, using the real data measured by the fast chirp FMCW-based high range resolution radar, we show that the proposed network is capable of learning the dynamics in time-series image data and outperforms the conventional classification schemes.

[1]  A. Cohen,et al.  GMM-based target classification for ground surveillance Doppler radar , 2006, IEEE Transactions on Aerospace and Electronic Systems.

[2]  Christopher Joseph Pal,et al.  Delving Deeper into Convolutional Networks for Learning Video Representations , 2015, ICLR.

[3]  Jürgen Schmidhuber,et al.  Learning Precise Timing with LSTM Recurrent Networks , 2003, J. Mach. Learn. Res..

[4]  Dit-Yan Yeung,et al.  Convolutional LSTM Network: A Machine Learning Approach for Precipitation Nowcasting , 2015, NIPS.

[5]  Christian Wolf,et al.  Action Classification in Soccer Videos with Long Short-Term Memory Recurrent Neural Networks , 2010, ICANN.

[6]  V. Winkler,et al.  Range Doppler detection for automotive FMCW radars , 2007, 2007 European Radar Conference.

[7]  Yoshua Bengio,et al.  Learning long-term dependencies with gradient descent is difficult , 1994, IEEE Trans. Neural Networks.

[8]  Yoshua Bengio,et al.  Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling , 2014, ArXiv.

[9]  Andrew Zisserman,et al.  Two-Stream Convolutional Networks for Action Recognition in Videos , 2014, NIPS.

[10]  Yoshua Bengio,et al.  On the Properties of Neural Machine Translation: Encoder–Decoder Approaches , 2014, SSST@EMNLP.

[11]  Wolfgang Menzel,et al.  High resolution automotive radar measurements of vulnerable road users – pedestrians & cyclists , 2015, 2015 IEEE MTT-S International Conference on Microwaves for Intelligent Mobility (ICMIM).

[12]  Akihiro Kajiwara,et al.  RCS measurements for vehicles and pedestrian at 26 and 79GHz , 2012, 2012 6th International Conference on Signal Processing and Communication Systems.

[13]  Jürgen Schmidhuber,et al.  Long Short-Term Memory , 1997, Neural Computation.

[14]  Andrew Gerald Stove,et al.  Linear FMCW radar techniques , 1992 .