LockedIn

About the Project

🧠 Inspiration

Our project was inspired by the challenges faced by individuals with ADHD and short attention spans, particularly when it comes to staying focused during study sessions. We wanted to create a tool that could provide real-time support and personalized feedback to help users manage their focus and productivity better.

💡 What it does

Our system uses an OpenBCI 8-channel EEG headset to monitor brain activity in real time. A custom-trained neural network analyzes the EEG data to determine the user’s concentration levels. Based on these levels, the system dynamically adjusts a study timer and provides recommendations—such as taking a break or continuing—by generating personalized feedback using Gemini API’s insight generation capabilities.

🛠 How we built it • Hardware: OpenBCI Cyton board (8-channel EEG headset) • Data Acquisition: EEG signals collected in real-time and processed into time windows • Machine Learning: Custom neural network trained to classify focus vs. non-focus states using preprocessed EEG features (like band powers, FFT outputs) • Integration: Gemini API was used to turn session data into insights and actionable advice • Frontend: A minimal UI to show the timer and concentration feedback • Backend: Python scripts for data streaming, signal preprocessing, ML inference, and insight generation

🧗 Challenges we ran into • We started with zero experience in EEG data handling, neural networks, and the Gemini API. • Setting up reliable data acquisition from the OpenBCI board was tricky due to noise and connection issues. • Training an accurate model required a lot of experimentation with signal preprocessing techniques and model architectures. • Integrating multiple systems (hardware, ML model, API) in real time without latency was a big hurdle.

🏆 Accomplishments that we’re proud of • Successfully built a real-time focus detection system using EEG data. • Trained a working neural network model that could classify user focus with reasonable accuracy. • Integrated the Gemini API to turn raw session results into personalized, human-readable insights. • Created a tool that could potentially help people with ADHD study more effectively.

📚 What we learned • How to work with brainwave data and the OpenBCI platform • Fundamentals of EEG signal processing and feature extraction • How to train and evaluate custom neural networks for classification • How to use external APIs (like Gemini) to enhance the usability and intelligence of an application • Real-world system integration with both hardware and AI models

🚀 What’s next for Untitled • Improving the accuracy of the concentration model with more labeled training data • Adding personalization features for different attention profiles • Creating a mobile app version for greater accessibility • Expanding the system to support neurofeedback-based learning