High Quality ECG-Codec

---

Introduction

This is the repo of implementation for ECG-Codec from paper: "Ultra-High Quality ECG Compression for IoMT Application Using Temporal Convolutional Auto-Encoder with Improved RVQ"

Features

Our ECG-Codec outperformed compared methods in terms of Quality Score(QS), which is a comprehensive metric evaluating performance of an ECG compressor. By maintaining low distortion while boosting our compress ratio, we achieved a QS of 42.7

Besides, we implemented three versions with different compression ratio of 44, 58 and 88, respectively.

To visualize our result intuitively, we showcase our reconstruction result on mit-bih datasets as below:

Version

first commit:2025.01.05 version 1

Citation

If your research have used our code, please cite this repo as below:

@misc{ECG-Codec,
  title={ECG-Codec: Ultra-High Quality ECG Compression Method},
  author={Yeyi Guan},
  year={2025},
  howpublished={\url{https://github.com/Guan-Y/ECG-Codec}},
}

Acknowledgements

The code is partially referred to:

• facebookresearch/encodec
• lucidrains/vector-quantize-pytorch