A new scheme of automatic modulation classification using Convolutional Neural Network with constellation diagram

抄録

Automatic Modulation Classification (AMC) of the transmitted signals remains a challenging area in modern intelligent communication systems such as cognitive radio system. In this research, we propose a new system for AMC using Convolutional Neural Network (CNN) with constellation diagrams in Additive White Gaussian Noise channel and use the trained model to recognite five digital modulation methods: BPSK, QPSK, 8PSK, 16QAM and 64QAM with constellation diagrams. We have achieved high efficiency and high precision recognition of modulation methods by using new CNN models, improved input image size, number of dropout layers and simulation hyperparameters.