Signal Processing and Speech Communication Laboratory
homeresults of the month › Complex-valued Convolutional Neural Networks for Enhanced Radar Signal Denoising and Interference Mitigation

Complex-valued Convolutional Neural Networks for Enhanced Radar Signal Denoising and Interference Mitigation

Published
Tue, Jun 01, 2021
Tags
rotm
Contact

Autonomous driving highly depends on capable sensors to perceive the environment and to deliver reliable information to the vehicles’ control systems. To increase its robustness, a diversified set of sensors is used, including radar sensors. Radar is a vital contribution of sensory information, providing high resolution range as well as velocity measurements. The increased use of radar sensors in road traffic introduces new challenges. As the so far unregulated frequency band becomes increasingly crowded, radar sensors suffer from mutual interference between multiple radar sensors. This interference must be mitigated in order to ensure a high and consistent detection sensitivity.

In this paper, we propose the use of Complex-valued Convolutional Neural Networks (CVCNNs) to address the issue of mutual interference between radar sensors. We extend previously developed methods to the complex domain in order to process radar data according to its physical characteristics. This not only increases data efficiency, but also improves the conservation of phase information during filtering, which is crucial for further processing, such as angle estimation. Our experiments show, that the use of CVCNNs increases data efficiency, speeds up network training and substantially improves the conservation of phase information during interference removal.

Figure: Empirical CDF performance comparison between the real-valued model (R-model, RVCNN), the complex-valued model (C-model, CVCNN), and the three classical methods zeroing, IMAT and Ramp filtering. Generally, the better the mitigation in terms of F1-Score, the higher is the EVM and thus the distortion of the phase at object peaks. In comparison to the RVCNN, the CVCNN reduces the peak distortions although it achieves a very similar F1-Score and it even outperforms the best classical method in terms of EVM, namely IMAT. Hence, the CVCNN achieves very high F1-Scores while also retaining low EVMs. The PPMSE shows similar characteristics. The CVCNN outperforms all other methods incorporating less phase distortions, particularly for weak interferences.

The full version of this paper can be found on https://arxiv.org/abs/2105.00929 and you can watch our conference presentation on https://youtu.be/u41e638zfMc.