https://github.com/jaredloomis/neurothink

Muse 2 EEG analysis with PyTorch + CNNs.
https://github.com/jaredloomis/neurothink

cnn-classification cnn-model eeg muse2 pytorch temporal-models

Last synced: 6 months ago
JSON representation

Muse 2 EEG analysis with PyTorch + CNNs.

• Host: GitHub
• URL: https://github.com/jaredloomis/neurothink
• Owner: jaredloomis
• Created: 2021-06-21T18:40:32.000Z (over 4 years ago)
• Default Branch: main
• Last Pushed: 2021-07-02T18:23:52.000Z (over 4 years ago)
• Last Synced: 2025-08-12T00:05:07.292Z (6 months ago)
• Topics: cnn-classification, cnn-model, eeg, muse2, pytorch, temporal-models
• Language: Jupyter Notebook
• Homepage:
• Size: 21.4 MB
• Stars: 14
• Watchers: 2
• Forks: 2
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# neurothink

neurothink is project exploring Muse 2 EEG data.

![EEG 10-10 system](./data/EEG-10-10.png)

## Eyes open/closed detection

[notebook](./EEG/Muse-EEG-eyes-open.ipynb)

In this notebook, I train a CNN to determine whether the wearer's eyes are open or closed based on the raw EEG signals. The results were surprising, with up to **82% accuracy** on my dataset.

Possible improvements:

- Use FFT data as additional features (ie. feature per band per sample). [example](https://www.kaggle.com/purplejester/pytorch-deep-time-series-classification)

- Use stationary wavelet transform, as a better (?) alternative to FFT.

- Experiment with network architectures.

## Direction detection

[notebook](./EEG/Muse-EEG-eyes-open.ipynb)

Here, I train a similar CNN to determine whether the wearer is thinking of the concept of `left`, `right`, or `neither`. **In Progress.**

## Meditation coach

Next, I plan on trying to training a network to detect different states of meditation, preferably with some notion of "meditation quality" baked into the training data and therefore the resulting network outputs.