EMI Threat Assessment of UAV Data Link Based on Multi-Task CNN

In this work, a multi-task convolutional neural network with multi-input (MIMT-CNN) is proposed for electromagnetic interference (EMI) signals recognition and environment risk evaluation of the data link unmanned aerial vehicle (UAV). The visualized performance parameters, short-time Fourier transform (STFT) spectrograms, constellation diagrams are obtained by experiment on susceptibility UAV’s datalink. particular, diagram further enhanced calculating density distribution sampling points to obtain normalized constellation. Taking above different categories images as input expected model, multi-element high correlation EMI features extracted fused in MIMT-CNN. Besides, structure series-parallel connection adopted trained model Bayesian optimization also used select hyperparameters. case, perception higher reliability can be obtained. On basis, complexity channels compared. results show that diagram, especially constellation, significantly improve accuracy model. Besides parameters STFT spectrogram inputs has much better performance.