10/19 Week Updates
For the week of October 19, the teams continued working on the project as the deadlines approach. There was not a significant amount of progress for most teams, rather there was more work on smaller details.
This week had the Hardware team discuss their journey to their present design. The team must include a 2 to 3 minute long segment in the project overview video that will be submitted to the NeurotechX competition by November 22nd. Apart from this, possible applications of the subvocalization headset were discussed. For example, detecting a language or assisting those with speech impediments.
- Cesar C. Ortega
This week, the Data team was able to increase the data accuracy to a stable 92%, an improvement from last week. The Data and ML team will continue to collaborate with each other to retrieve more data. As of now, the data set is at 416, which is around 14 minutes of recordings. As the Data team receives more data from a new person's vocalization, they will have to be cautious about how this data will affect their current accuracy. -Aracely Cano-Gramajo
Machine Learning Team
This week no data came so the model couldn’t be tuned. But, the Data team pipeline was tuned to reach 92% accuracy. The Data and ML team will continue to collaborate to make the best model using the new data coming next week hopefully, and try to maintain the same accuracy by November 8. By November 15, the ML and Data team will try to get their videos done as well. -Brandon Luu
The UI team was able to display the hardware port number on the interface, a problem they had last week. Productivity within the team has been better this week. For the upcoming week, the UI team plans to meet with the Hardware team to finalize the demonstration of the hardware and interface. -Aracely
- 7 toasts