Imagined Speech BCI
Real-time imagined-speech BCI with NeuroPawn EEG at NATHacks 2025.
How It Works
Unlike many BCI demos that rely on motor imagery or subtle muscle movements, this project focused on true imagined speech: detecting which word a user was thinking of without speaking, moving, or subvocalizing.
The team trained a model to classify EEG patterns across three imagined words, along with noise, and achieved a working real-time demo with approximately 72% accuracy.
Why It’s Hard
This project highlights one of the most challenging areas of BCI research. EEG signals are highly variable, noisy, and difficult to interpret, especially when the target is internal thought rather than physical movement. Despite those challenges, the team was able to build a functional prototype in just 72 hours.
At NeuroPawn, we are proud to support ambitious student projects that push the boundaries of what accessible EEG hardware can be used for. Projects like this show how hands-on neurotechnology can help students explore difficult research problems and turn bold ideas into working demonstrations.
