Robustness of deep neural networks for micro-Doppler radar classification

Czerkawski, Mikolaj and Clemente, Carmine and Michie, Craig and Andonovic, Ivan and Tachtatzis, Christos (2022) Robustness of deep neural networks for micro-Doppler radar classification. In: International Radar Symposium 2022, 2022-09-12 - 2022-09-14. (In Press)

[thumbnail of Czerkawski-etal-IRS-2022-Robustness-of-deep-neural-networks-for-micro-Doppler-radar-classification] Text. Filename: Czerkawski_etal_IRS_2022_Robustness_of_deep_neural_networks_for_micro_Doppler_radar_classification.pdf
Accepted Author Manuscript
Restricted to Repository staff only until 14 September 2022.
License: Strathprints license 1.0

Download (1MB) | Request a copy

Abstract

With the great capabilities of deep classifiers for radar data processing come the risks of learning dataset-specific features that do not generalize well. In this work, the robustness of two deep convolutional architectures, trained and tested on the same data, is evaluated. When standard training practice is followed, both classifiers exhibit sensitivity to subtle temporal shifts of the input representation, an augmentation that carries minimal semantic content. Furthermore, the models are extremely susceptible to adversarial examples. Both small temporal shifts and adversarial examples are a result of a model overfitting on features that do not generalize well. As a remedy, it is shown that training on adversarial examples and temporally augmented samples can reduce this effect and lead to models that generalise better. Finally, models operating on cadence-velocity diagram representation rather than Doppler-time are demonstrated to be naturally more immune to adversarial examples.

ORCID iDs

Czerkawski, Mikolaj ORCID logoORCID: https://orcid.org/0000-0002-0927-0416, Clemente, Carmine ORCID logoORCID: https://orcid.org/0000-0002-6665-693X, Michie, Craig ORCID logoORCID: https://orcid.org/0000-0001-5132-4572, Andonovic, Ivan and Tachtatzis, Christos ORCID logoORCID: https://orcid.org/0000-0001-9150-6805;