Automatic Modulation Classification in the Presence of Interference
Pavlos Triantaris, Evgeny Tsimbalo, Woon Hau Chin, Denız Gündüz · 2019
A modulation recognition method based on a convolutional neural network (CNN) architecture is assessed through classification of synthetic baseband signals in the presence of a second interfering signal source. The complexity and adaptability of CNNs is leveraged so as to forgo statistical feature extraction procedures and efficiently classify based on raw signals or their modified forms. Both scenarios with the interfering signal's modulation scheme known and unknown, are considered. Simulation results show that the CNN architecture achieves considerable accuracy despite the presence of interference, and the knowledge of the modulation scheme of the interfering signal significantly improves the accuracy.