.png){kind=icon}
.png){kind=icon}
Return
SMTrans: An efficient automatic modulation recognition network based on the scale-aware modulation transformer
.png){kind=icon}
PRE
AI
DOI：10.1016/j.phycom.2025.102966.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In recent years, with the rapid advancement of deep learning technologies, automatic modulation recognition, as a crucial component of blind signal processing, has attracted extensive attention from the research community. However, most deep learning-based AMR models tend to focus excessively on recognition accuracy while neglecting computational efficiency, posing significant challenges for deployment on embedded and edge devices. In this paper, we propose an efficient automatic modulation recognition network based on a scale-aware modulation unit, which integrates the advantages of convolutional neural networks and Transformers. The proposed architecture effectively reduces both the number of parameters and computational complexity while maintaining high recognition accuracy. Experimental results demonstrate that the proposed model achieves recognition accuracies of 63.27% and 65.17% on the RadioML2016.10a and RadioML2016.10b datasets, respectively. Its performance is comparable to state-of-the-art models that require five times more parameters and fifteen times more FLOPs, while also surpassing existing lightweight AMR models in terms of recognition accuracy. Furthermore, to enhance the model's performance, we conducted an in-depth investigation into the impact of various data augmentation sstrategies, leading to an additional 0.55 % improvement in recognition accuracy.
Keywords:
Automatic modulation recognition
Deep learning
Convolutional neural networks
Transformers
Data augmentation
Journal
IF:
2.2
Papers:
360
Citations:
2.6K