Inspiration
This project was inspired by a curiosity in Neuroscience and EEG data. When reviewing brain states and HRK/stress data, we thought it might be possible to maximize someone's efficiency and reduce their stress by limiting the amount of time they spend on certain media applications during high stress periods and providing useful insights with the user's long-term data as to which programs can cause or reduce high stress states.
What it does
The project analyzes HRK and Heart Rate data sourced from an Apple Watch as well as optional EEG data sourced from an EEG headband to analyze brain state and stress level constantly and limit select applications (user-selected) during high-stress periods. It also collects user data to provide context to Gemini over a period of time using an MCP server, so that Gemini can analyze and inform the user of specific applications that cause them stress during use and inform them of their good habits. This also allows for an active fluctuation of what we should consider as someone's base HRK and Heart Rate throughout different parts of the day.
How we built it
This project was developed as a Swift Application that starts a custom workout using the Apple HealthKit API to constantly monitor Heart rate and HRK, and PyLSL to actively transfer the EEG data. At a basic level, it then analyzes the information and restricts applications if HRK falls under a baseline level, which will be later adapted as decided by our Gemini model. The EEG headset provides further insight to the user.
Challenges we ran into
Apple HealthKit and other IOS APIs are bad :( Also, we did not have hardware such as an Apple Watch 4+ (with ECG data) to test with.
Accomplishments that we're proud of
The application is functional and properly tracks data :)
What we learned
We learned IOS application development and how to get around IOS app/API restrictions. Also, this project required a lot of research on heart health data and neuroscience/EEGs.
What's next for NeuroFade
Next is full scale deployment and testing with actual hardware. Once real-time data is being collected, our model will be able to train properly and analyze the HealthKit/EEG data. It is also possible that we add more information and insight from other collected health info to improve accuracy.
Log in or sign up for Devpost to join the conversation.