9/7 Week Updates
Currently the hardware team is working out some kinks with wiring and placement of the electrodes. A main goal is to have something printed to see if things fit together properly. Firstly, they would like to print the encasing for the board and have it be separate from the batteries. This allows for more efficient use of the 3D printing materials since the board casing may need to go through multiple iterations as the headband changes. Looking towards the future, more recordings that are 2 minutes in length need to be completed and we might leave open BCI for a new software.
This week the data team is almost done with building the pipeline and everyone has moved to processing. They also got everyone on github. For processing, they are learning and implementing things such as normalization, butterworth filter, ICA to filter the data. For now, the butterworth filter is finished. This is the basic pipeline for the data team. Receive ~2 minute recording file from hardware team > splice into the 2 second clips based on lap > bandpass/butterworth filter > ICA > normalization > output into pkl file or something aligned with label (also based on lap)
Machine Learning Team
Machine Learning is continuing on with last weeks plans. The models previously developed are being improved. They are also striving to get their 1 dimensional convolutions data working. Other than that, they are prepping for the deep learning workshop they are hosting on Wednesday.
The UI team is adding a mechanism for providing feedback to train the machine learning model. Questions will be displayed with a Yes/No prompt so the user can provide the expected output. Then the result is passed to the model and if it guessed incorrectly, it will automatically adjust itself to try and be more accurate in the future.
- 8 toasts