Repository files navigation

Official Code for "IQFormer: A Novel Transformer-Based Model With Multi-Modality Fusion for Automatic Modulation Recognition". [paper]

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={2024},volume={},number={},pages={1-1},keywords={Feature extraction;Modulation;Transformers;Convolution;Time-frequency analysis;Wireless communication;Time-domain analysis;Signal to noise ratio;Market research;Interference;Automatic modulation recognition;deep learning;multi-modality fusion;transformer},doi={10.1109/TCCN.2024.3485118}}

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 toRichardzhangxx 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.

withh5py.File(os.path.join(args.database_path,'HisarMod2019train.h5'))ash5file:train=h5file['samples'][:]train_label=h5file['labels'][:]SNR_tr=h5file['snr'][:]h5file.close()withh5py.File(os.path.join(args.database_path,'HisarMod2019test.h5'))ash5file: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

pythonmain.py

If you want our pre-trained models on both three datasets, please contactshaomy666@stu.xmu.edu.cn

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

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.