Domain Adaptation for Deep Learning based Signal Detection in the High Frequency Range

Ciritci Y, Buhl U, Fischer G (2025)

Publication Language: English

Publication Type: Conference contribution, Conference Contribution

Publication year: 2025

Pages Range: 1992-1996

Conference Proceedings Title: 33rd European Signal Processing Conference EUSIPCO 2025

Event location: Isola delle Femmine – Palermo IT

ISBN: 978-9-46-459362-4

Abstract

Signal detection, especially in crowded frequency ranges, poses high challenges to any signal analysis solution. Therefore, in recent years, approaches apply deep learning object detection methods to the specific signal detection task. Although in principle huge similarities between object detection and signal detection exist, applying specific object detection solutions is not always straightforward though. In this work, we focus on the specific challenge of domain adaptation for deep learning signal detection. We adapt ideas from domain adaptation for object detection and introduce a novel method called Domain Adaptation with Multi-Class domain classifiers (DA-MC). This method surpasses the performance of domain adaptation with class-agnostic domain classifiers. The latter prove to be inapplicable for signal detection domain adaptation. We train and evaluate our proposed method on a synthetic test scenario. Our results prove that the domain shift for signals can effectively be solved. We furthermore evaluate an actual domain shift signal scenario using an over-the-air dataset and present promising improvements over a baseline training. 

Authors with CRIS profile

Yasin Ciritci Professur für Technische Elektronik Georg Fischer Professur für Technische Elektronik

Involved external institutions

Rohde & Schwarz GmbH & Co. KG DE Germany (DE)

How to cite

APA:

Ciritci, Y., Buhl, U., & Fischer, G. (2025). Domain Adaptation for Deep Learning based Signal Detection in the High Frequency Range. In 33rd European Signal Processing Conference EUSIPCO 2025 (pp. 1992-1996). Isola delle Femmine – Palermo, IT.

MLA:

Ciritci, Yasin, Ulrike Buhl, and Georg Fischer. "Domain Adaptation for Deep Learning based Signal Detection in the High Frequency Range." Proceedings of the EUSIPCO 2025, Isola delle Femmine – Palermo 2025. 1992-1996.

BibTeX: Download