Overview
- Built a P300 BCI decoder in Python fusing EEG neural signals with a character-level language model prior in order to reduce the total number of flash repetitions needed for accurate character selection from the speller interface.
- Engineered ML pipeline with ERP epoch extraction, regularized LDA with Platt scaling, and Bayesian fusion across 200+ hours of EEG data from 36 participants, targeting faster communication for ALS patients.