9/28 Week Updates
As the fall quarter begins and teams figure out their fall schedule the first prototype (image bellow) of the headset has been printed! Some issues with this iteration were that the headset was too wide and the discs had to be cut short from the bottom to lay flat on the 3D printer tray. The headset opening was originally 170mm wide thus the next print will have the opening be about 156mm. Lastly, padding was added for now to aid in the recording process.
-Cesar C. Ortega
This week the data team met with the machine learning team and fixed the labeling for alternating files, which caused problems with the machine learning model. One problem they are trying to solve is that too much data is being removed when it is being processed. This is mainly because of the Independent Component Analysis(ICA) taking off too much data. As of right now, the precision level is around 80%, but the goal is 95%. To fix this problem, they are looking to make the Butterworth filter weaker. Then if that doesn’t work taking out the ICA, then trying to remove the normalization as a last resort.
Machine Learning Team
The majority of the Machine Learning team’s past meeting was spent working together on various aspects of their projects, but they had a few moments of discussion in the beginning. They started off by talking about future meeting times that work for everyone with classes coming up. After that, they recounted some errors they were having trouble with in their tests, and concluded that a miscommunication in their filing convention was causing some confusion.
The UI team added a prompt for the user to input whether or not the ML model's guess was accurate. They also added additional information to the main display and cleaned up the interface.
- 10 toasts