Research Article
Research on Several Neural Network Structure for Automatic Modulation Recognition
@INPROCEEDINGS{10.4108/eai.21-11-2024.2354628, author={Yidong Xu}, title={Research on Several Neural Network Structure for Automatic Modulation Recognition}, proceedings={Proceedings of the 2nd International Conference on Machine Learning and Automation, CONF-MLA 2024, November 21, 2024, Adana, Turkey}, publisher={EAI}, proceedings_a={CONF-MLA}, year={2025}, month={3}, keywords={automatic modulation recognition deep learning convolutional neural network residual neural network}, doi={10.4108/eai.21-11-2024.2354628} }- Yidong Xu
Year: 2025
Research on Several Neural Network Structure for Automatic Modulation Recognition
CONF-MLA
EAI
DOI: 10.4108/eai.21-11-2024.2354628
Abstract
With the rapid development of communication technology, Automatic Modulation Recognition (AMR) based on Deep learning (DL) performs well relying on its unique advantages. However, due to the wide variety of neural networks, it is important to compare and analyze their performance and applicability under specific conditions. In this paper, we select convolutional neural network (CNN) and Residual networks (Resnet), and continuously deepen the depth of the residual network to explore the influence of the accumulation of residual blocks. After simulating and analyzing the recognition effects of different network structures under -12 to 30 signal-to-noise ratio (SNR) conditions, the experimental results show that under the experimental conditions set up in this paper, the recognition rate of Resnet is about 4.8% higher than that of CNN on average when SNR is higher than 0db. After accumulating one and two residual blocks and fine-tuning the model to improve the recognition rate, the recognition rate of both networks obtained from the improvement exceeds 90% when SNR is higher than 10db.
