A Hermitian Positive Definite neural network for micro-Doppler complex covariance processing

Daniel A Brooks (1, 2) , Olivier Schwander (2) , Frédéric Barbaresco (3) , Jean-Yves Schneider , Matthieu Cord (2)

1 Thales LAS France
2 MLIA - Machine Learning and Information Access
3 Thales Air Systems

In its raw form, micro-Doppler radar data takes the form of a complex time-series, which can be seen as multiple realizations of a Gaussian process. As such, a complex covariance matrix constitutes a viable and synthetic representation of such data. In this paper, we introduce a neural network on Hermitian Positive Definite (HPD) matrices, that is complex-valued Symmetric Positive Definite (SPD) matrices, or complex covariance matrices. We validate this new architecture on synthetic data, comparing against previous similar methods.

Mots clés: drone classification; micro- Doppler; complex numbers; neural networks; covariance matrices

Domaines: Informatique [cs]; Intelligence artificielle [cs.AI]; Réseau de neurones [cs.NE]

Daniel A Brooks, Olivier Schwander, Frédéric Barbaresco, Jean-Yves Schneider, Matthieu Cord. A Hermitian Positive Definite neural network for micro-Doppler complex covariance processing. International Radar Conference, Sep 2019, Toulon, France. ⟨hal-02422456⟩