Overview of Automatic Modulation Recognition Methods based on Deep Learning

Spectral resources are becoming increasingly scarce due to the rapid growth of communication data volumes, and the modulation methods of signals will also become more diverse. There is an urgent need for more effective modulation recognition methods. Firstly, based on a large amount of high-level information, this paper introduces shallow machine learning and some of its representative methods. Secondly, in response to the low accuracy and high computational complexity of most shallow machine learning models, this paper introduces the application of deep learning in the field of automatic modulation recognition. Finally, possible solutions to the existing problems and challenges in modulation recognition technology are proposed, and future prospects are discussed. Automatic Modulation Recognition(AMR) technology is relatively well developed in closed-set research, but it still needs to be strengthened in open-set research. In addition, research on unlabeled samples is also lacking, and there should be an increased focus on research for powerful hardware platforms capable of modulation recognition.