Modulation Recognition Based on Lightweight Neural Networks

Abstract

In order to solve the problems of complex networks, large amount of calculation and high equipment requirements in the current deep learning method to complete the modulation recognition process, this paper proposes a modulation recognition algorithm based on lightweight neural networks. First, map the common 8 kinds of modulated signals to constellation diagrams to make image data sets. In the process of retaining the original signals, make full use of the performance of the neural network, build a representative the MobileNet neural network in the neural network to complete the training of the data set, use the test samples Verify the effectiveness of the lightweight neural networks used. Simulation experiment results show that the overall recognition rate of modulation reaches 98% when the SNR is greater than 2dB, but the training speed is greatly improved.