Profes Tackle Sleep

We did the Neurotech track, which involved taking EEG and other data and using it to label what sleep stage the person is in. A lot of out time and effort went in to feature engineering, where we used Fourier transforms, along with advanced filtering and smoothing techniques, to transform raw brain waves data into a vector that shows the power of the discrete frequencies of delta, theta, alpha, beta, and gamma waves. The frequency domain for these signals more accurately correlates with sleep stages, and both the final "block" model and temporal "sub-epoch" model perform at over 90% accuracy on test data.

Built With

• matplotlib
• mne
• python
• scipy
• tensorflow