Develop SpiroMask end-to-end.

Authors

• Vanshika Jain Sophomore in Electrical Engineering at NIT Rourkela.

Navigation and Installation

You can do the below installations or just view the files from the repository directly. Both the python file(SRIPTask-2.ipynb) and the quarto file(SRIPTask-2-2.pdf) are present.

1. Clone the repository:

   git clone https://github.com/26vanshika/SRIP24.git
   cd SRIP24
   

2. Set up a Virtual Environment

    python -m venv venv
    # For mac
    source venv/bin/activate
    # For Windows
    use ".\venv\Scripts\activate"
   

3. Install Dependencies

   pip install -r requirements.txt
   

4. Run Jupyter Notebook

   jupyter notebook
   

5. Run my file In your web browser, go to the URL displayed in the terminal http://localhost:8888/ . Open main Jupyter Notebook file SRIPTask-2.ipynb.

6. Quarto file

   quarto render SRIPTask-2.ipynb --to pdf
   

Roadmap

The steps that I followed for this mini-project were as follows:

I first gathered and learnt about the resources I would be needing by searching them through the internet.

It was not a one day work though the project seems small. I ran into multiple challenges which I will be mentioning below.

I first recorded the audio of breathing in mask.

Then using IPython I imported it in my jupyter notebook.

To do the analysis of the audio I used Librosa package mainly used for Sound analysis.

I learnt about the low-pass filter to filter high frequencies and coded one(I faced a lot of issues here :) ).

Then using Matplotlib and Librosa's display feature I plotted the waveform and spectogram of Original and Filtered audio.

What I learned

So basically I had pre-requisite knowledge in Machine learning, but I learnt a lot of things

Like I was aware of Matplotlib but wasnt aware of Librosa's waveform feature.

I also learnt how to make filters in python(I only learnt about them in theory thanks to an Electrical Engineering course)

Also Quarto is just something I didnt know and glad through this project I learnt about it, because it really does wonders. Definetly going to use in my future projects.

Challenges I ran into

There were a lot of challenges I ran into:

Firstly my packages were installed but were showing "module not found", fixed that with a lot of frustration.

I also faced issues with Quarto since I was completely new to it.

And a lot of errors in making the low-pass filter.

My github was also not committing changes smoothly, it was continously showing errors.

References

https://www.comet.com/site/blog/working-with-audio-data-for-machine-learning-in-python/

https://quarto.org/docs/tools/jupyter-lab.html

https://medium.com/analytics-vidhya/how-to-filter-noise-with-a-low-pass-filter-python-885223e5e9b7

https://stackoverflow.com/questions/25191620/creating-lowpass-filter-in-scipy-understanding-methods-and-units