Skip to content

Latest commit

 

History

History
80 lines (55 loc) · 3.13 KB

README.md

File metadata and controls

80 lines (55 loc) · 3.13 KB

MAMC

A method for Automatic Modulation Classification using Mamba structure

A project employing the Selective State Space Model (Mamba) method for Automatic Modulation Classification (AMC) in a scenario of extended signal length.

The increased sequence length complicates the learning process and diminishes accuracy, while simultaneously escalating memories and reduces timeliness. This issue brings the following impacts: Length effects

if our codes helped your reasearch, please consider citing the corresponding submission

arXiv

@article{zhang2024MAMC,
  title={MAMCA -- Optimal on Accuracy and Efficiency for Automatic Modulation Classification with Extended Signal Length},
  author={Yezhuo Zhang and Zinan Zhou and Yichao Cao and Guangyu Li and Xuanpeng Li},
  year={2024},
  journal={arXiv preprint arXiv:2405.11263},
}

IEEE Communications Letters (Early Access)

MAMC

@article{10705364,
  author={Zhang, Yezhuo and Zhou, Zinan and Cao, Yichao and Li, Guangyu and Li, Xuanpeng},
  journal={IEEE Communications Letters},
  title={MAMC - Optimal on Accuracy and Efficiency for Automatic Modulation Classification with Extended Signal Length},
  year={2024},
  volume={},
  number={},
  pages={1-1},
  doi={10.1109/LCOMM.2024.3474519}
}

We utilize a denosing unit for better accuracy performance under noise interference, while using Mamba as the backbone for low GPU occupancy and training/inference time.

  • MAMC structure MAMC_structure

To related AMC works, as well as sorce code:

  • Deep Learning Based Automatic Modulation Recognition: Models, Datasets, and Challenges

    AMR-Benchmark

To the denosing method employed in our work, as well as source code:

To the Mamba method employed in our work, as well as source code:

Requirements

pip install -r requirements.txt

Training

cd into code/script and do

bash RML2016.10a.sh

Contact

If you have any problem with our code or any suggestions, including discussion on SEI, please feel free to contact