Modulation Pattern Recognition Based on Resnet50 Neural Network
Xiao Tian, Chao Chen · 2019
In this paper, several typical convolutional neural networks, VGG16, VGG19, Inception, Xception and Resnet50, are used to identify the modulation pattern of the constellation. It is found through experiments that the Resnet50 network performs best in the recognition of constellations. Then, using the Resnet50 convolutional neural network, the two modes of the modulation mode and the signal-to-noise ratio of the constellation are simultaneously identified. When the accuracy of the signal-to-noise ratio is required to be 1 db, the identified quasi-group rate is 60 percent. When the accuracy of the signal to noise ratio is required to be 2db. The accuracy of recognition can reach 85 percent.