| Autor: |
Horn, Claus, Nyfeler, Matthias, Müller, Georg, Schüpbach, Christof |
| Jazyk: |
angličtina |
| Rok vydání: |
2022 |
| Předmět: |
|
| DOI: |
10.21256/zhaw-27185 |
| Popis: |
We develop a multi-timescale deep learning algorithm to detect drones from radio signals. While previous approaches focused on the analysis of high-frequency radio data alone we integrate signals from the higher timescale of the drone communication protocol in an end-to-end architecture. To this end, we develop a new meta-CNN layer, which generalizes the idea of the standard CNN (which slides a single, fully connected kernel along a higher-level input) towards arbitrarily complex kernel models. To detect higher timescale patterns our system uses an LSTM layer in the top layers. As a result, our model is able to extend drone identification abilities significantly toward very small SNRs. |
| Databáze: |
OpenAIRE |
| Externí odkaz: |
|
