Robustness of Automatic Modulation Classification Based on a Deep Neural Network in the Presence of Interference

Automatic modulation classification is an essential functionality of many state-of-the-art communications technologies. We study the robustness of automatic modulation classification based on a convolutional neural network to interference. In our scenario, the modulation of the interference signal is unknown and is not included in the set of modulations on which the neural network is trained and validated. We use detailed models of the communication channel and hardware imperfections for all our simulations. Our results demonstrate the high robustness of the convolutional neural network to interference in the studied situations.