We Use Muse

Inspiration

We wanted to explore what can be achieved in terms of thought-based computer control using the Muse Headband.

What it does

One can use the headband's accelerometer to control the cursor, but the thought pattern classification subsystem does not work very well, as it is very hard to collect clean training data without a pure lab environment (e.g. no distractions, etc.).

We also built a simple prototype of an interface which would allow a user to track how focused they are during their work. If someone were to, perhaps, spend a long time working without being productive, this app could provide insights into why this was the case: it would highlight cases when a user would have trouble focusing.

How we built it

We used Python and Ruby, as well as the Muse research tools. For the data analysis, we used Jupyter and scikit-learn. For the perfect UX and design, we leveraged Comic Sans MS™ and a patented marquee implementation.

Challenges we ran into

It's very difficult to analyze and classify traces of EEG activity in such a limited timeframe. Moreover, collecting clean data with no distractions is nearly impossible in hackathon conditions.

Accomplishments that we're proud of

We collected quite a lot of data of people performing various activities while wearing the Muse headband. The accelerometer-based computer control interface also works quite well!

What we learned

EEG-powered human-computer interaction is possible, i.e., the signal does exist in the data. With appropriate training data collection and analysis, one could definitely build a real-time computer control system using EEG.

What's next for We Use Muse

We would like to improve our data collection conditions, and build a more robust pipeline for the data cleaning.

Built With

• comic-sans
• muse
• muse-headband
• python
• ruby
• scikit-learn