This repo contains notebooks working through an example of deconvoluting mixed audio signals using Independent Component Analysis (ICA) in sklearn. It synthetically mixes audio files in generate_example_mixes.ipynb. It then deconvolutes the mixed signals using ICA and compares them to the original individual audio files in deconvolute_example_mixes.ipynb. Functions useful for manipulating audio files are included in wav_utils.py.
Deconvoluted audio clip (cello only)
- Udacity: Introduction to Machine Learning with TensorFlow
- This notebook is a deeper exploration inspired by an example from Udacity
- Blind source separation using FastICA
- Simple example on sklearn
# set up virtual env with pyenv
pyenv virtualenv 3.11.7 ica-example
pyenv shell ica-example
# clone the repo
git clone git@github.com:mrperkett/ICA-audio-deconvolution.git
cd ICA-audio-deconvolution/
# install requirements
python3 -m pip install requirements.txt
# register IPython kernel
python3 -m ipykernel install --user --name ica-example
Run Jupyter lab, open the notebook you wish to run, and select the ica-example kernel to run.