Drone Classification with a Convolutional Neural Network Applied to Raw IQ Data
Stefan Kunze, B. Saha · 2022 3rd URSI Atlantic and Asia Pacific Radio Science Meeting (AT-AP-RASC) · 2022
With the increasing popularity of civilian drones, the need for technical detection and classification systems rises. In this paper a machine learning based approach for detection and classification of radio frequency signals from drones is proposed. As data source the DroneDetect_V2 data set is used. The raw IQ data is processed by a convolutional neural network, without the need for much pre-processeing or any feature engineering. With this approach an accuracy of 99 % for detection and between 72 % and 94 % for classifi-cation is reached.