Modulation Recognition Analysis Based on Neural Networks and Improved Model

Abstract

With the emergence and development of computer technology, software technology and other technologies, there are a number of new solutions to the problem of modulation recognition. Modulation recognition is a key step in digital signal processing. The traditional recognition method based on likelihood ratio decision theory has gradually been replaced by the recognition method based on neural network. In this paper, for the recognition of modulation signals, in the PyTorch framework, Convolutional Neural Networks and Residual Neural Network are used to identify modulation signals and test the accuracy. In addition, the advantages and disadvantages of the two models for modulation signal recognition performance are compared, and the model is improved at the same time. In this paper, linear layers with different parameters are added to the Convolutional Neural Networks to observe the impact of its recognition rate and loss function, and a new improvement direction for the recognition of modulation signal datasets is proposed.