Neural Network Models for Detection of Unmanned Aerial Vehicles Based on Spectrogram Analysis

The widespread use of unmanned aerial vehicles (UAVs) in both the national and military fields has made them the focus of industrial organizations. However, the use of UAVs has seriously affected the violation of the confidentiality of personal data, posed a threat to states, national institutions, nuclear power plants, historical places. One way to reduce this threat is to detect harmful UAVs. The article develops machine learning and deep learning methods based on sound signal analysis to detect harmful UAVs. Features were extracted from the sound signals and their ensemble was created. The created new data was transmitted to the input of neural network models in the form of vectors and drones were detected. The effectiveness of the proposed approach has been tested on a database open to scientific research.