IQFormer: A Novel Transformer-Based Model With Multi-Modality Fusion for Automatic Modulation Recognition
Official Code for “IQFormer: A Novel Transformer-Based Model With Multi-Modality Fusion for Automatic Modulation Recognition”. [paper]
Citation
If our work is helpful to your research, please star us on github and cite :
@ARTICLE{10729886,
author={Shao, Mingyuan and Li, Dingzhao and Hong, Shaohua and Qi, Jie and Sun, Haixin},
journal={IEEE Transactions on Cognitive Communications and Networking},
title={IQFormer: A Novel Transformer-Based Model With Multi-Modality Fusion for Automatic Modulation Recognition},
year={2025},
volume={11},
number={3},
pages={1623-1634},}
Preparation
We conducted experiments on three datasets, namely RML2016.10a, RML2016.10b and HisarMod2019.1.
The datasets can be downloaded from the DeepSig(RML2016 series), HisarMod2019.1. Special thanks to Richardzhangxx for providing the .MAT file for HisarMod2019.1. For your convenience, I have combined the I/Q signals and saved them in the h5py file. If you want to know the most widely used dataset division ratio, read my paper.
with h5py.File(os.path.join(args.database_path, 'HisarMod2019train.h5')) as h5file:
train = h5file['samples'][:]
train_label = h5file['labels'][:]
SNR_tr = h5file['snr'][:]
h5file.close()
with h5py.File(os.path.join(args.database_path, 'HisarMod2019test.h5')) as h5file:
test = h5file['samples'][:]
test_label = h5file['labels'][:]
SNR_te = h5file['snr'][:]
h5file.close()
Please extract the downloaded compressed file directly into the ./dataset directory, or change the
args.database_path. args.database_choose should keep in [2016.10a, 2016.10b, 2019].
Then you just need to
python main.py
If you want our pre-trained models on both three datasets, please contact [email protected] or download link: https://pan.quark.cn/s/898814b98b7e
Environment
These models are implemented in Keras, and the environment setting is:
- Python 3.11
- Pytorch 1.12.0
- pandas
- seaborn
- h5py
- scikit-learn
- matplotlib
- tensorboardX
- tqdm
- timm
License
This code is distributed under an MIT LICENSE. Note that our code depends on other libraries and datasets which each have their own respective licenses that must also be followed.
评论