Deep learning based higher-order approximation for multiple knife edge diffraction

Viet-Dung Nguyen (1, 2) , Huy Phan , Ali Mansour (1, 2) , Arnaud Coatanhay (1, 3) , Thierry Marsault

1 ENSTA Bretagne - École Nationale Supérieure de Techniques Avancées Bretagne
2 Lab-STICC_SI3 - Equipe Security, Intelligence and Integrity of Information
3 Lab-STICC_PIM - Equipe PIM

We introduce an hybrid approach for computing multiple knife-edge diffraction attenuation. First, we show that the well-known Epstein-Peterson method can be considered as the first order approximation of the Vogler method. In the other words, the Vogler method is a combination of the Epstein-Peterson method and an higher order approximation. Then, we propose to learn the approximation based on deep learning methods. The key advantage of this approach is the significant reduction of generating training data to approximate the Vogler method while still offering a good accuracy and fast computation. Comparison to the state-of-the-art methods demonstrates the effectiveness of our proposed approach.

Mots clés: Deep learning; Learning systems

Domaines: Traitement du signal et de l'image [eess.SP]

Viet-Dung Nguyen, Huy Phan, Ali Mansour, Arnaud Coatanhay, Thierry Marsault. Deep learning based higher-order approximation for multiple knife edge diffraction. 2022 IEEE International Symposium on Antennas and Propagation and USNC-URSI Radio Science Meeting (AP-S/USNC-URSI), Jul 2022, Denver, United States. pp.1960-1961, ⟨10.1109/AP-S/USNC-URSI47032.2022.9886567⟩. ⟨hal-03839164⟩