LingoNMF

Non-negative Matrix Factorization (NMF) Enhanced by Large Language Models (LLMs) For Cardiorespiratory Sound Separation

How to Run the Code

Run the LingoNMF Jupyter Notebook
Open Jupyter Notebook and run LingoNMF.ipynb to execute the code. Open Jupyter Notebook and run LLama.ipynb to execute the LLM prompts (GPU needed).

Or:

1. Install Dependencies
   Install the required Python packages using:

   pip install -r requirements.txt
   

2. Run the Main Python Script
   Execute the main script using:

   python main.py
   

3. Run the MATLAB implementation

   Execute PL_NMF.m for NMF implementation in MATLAB

Citation:

If you use this code in your research, please cite:

• Torabi, Yasaman; Shirani, Shahram; Reilly, James P. (2025), Large Language Model-based Nonnegative Matrix Factorization For Cardiorespiratory Sound Separation, arXiv preprint, https://doi.org/10.48550/arXiv.2502.05757.
• Torabi, Yasaman; Shirani, Shahram; Reilly, James P. (2023), A New Non-Negative Matrix Factorization Approach for Blind Source Separation of Cardiovascular and Respiratory Sound Based on the Periodicity of Heart and Lung Function, arXiv preprint, https://doi.org/10.48550/arXiv.2305.01889.

If you use this dataset in your research, please cite:

• Torabi, Yasaman; Shirani, Shahram; Reilly, James P. (2024), Manikin-Recorded Cardiopulmonary Sounds Dataset Using Digital Stethoscope, arXiv preprint, https://doi.org/10.48550/arXiv.2410.03280

© 2025 by Yasaman Torabi. All rights reserved.