Modulation Classification Based on Signal Constellation and Yolov3

Abstract

Aiming at the modulation classification of communication signal, a new identification method based on signal constellation is proposed. In this method, the constellation of communication signal is identified as an image, and the Y olov3 based on deep learning is used for target recognition. This method is not limited by the order of signal constellation, and has the advantage of not requiring manual feature selections. Firstly, the constellation diagrams of different modulation signals are produced and then they are used for training the weight of yolov3. Finally, the testing image of received signal constellation is detected by yolov3 under different noises. Simulation results demonstrate that the proposed method can identify more kinds of signals and has better performance than the traditional cumulant-based method.