Finger Actuation – BCI

This project was conducted as part of a competition. We were given 3 data-sets, for 3 different subjects. All 3 subjects were given an experiment wherein they had to move their fingers to different positions. The 256-channel ECOG data for this was recorded. Using these data sets, we had to decode 3 ECOG datasets to find out what the actuation of the fingers were.

The approach to this project involved a lot of Machine Learning. We started using basic linear models, with feature selection. The results, however, were primarily inaccurate for this method.

In the end, we used a CNN model to decode the signals.