Simulated online typing performance in a cBCI using different language models

Tab Memmott, Dylan Gaines, Matthew Lawhead, Dan Klee, Barry Oken, Keith Vertanen

Proceedings of the 11th International Brain-Computer Interface Meeting, 2025.

Communication Brain-Computer Interfaces (cBCIs) represent a crucial technological advancement for individuals with severe motor disabilities as they offer a direct pathway to express their thoughts and needs without physical movement. These systems commonly leverage the P300 ERP, a distinct neural response approximately 300-500ms after a novel stimulus. Language modeling presents a promising approach to enhancing the performance and usability of cBCIs. However, integrating language models with cBCI systems presents unique challenges, including balancing model complexity with real-time processing requirements and optimizing system performance parameters. This study utilizes simulations of online cBCI data to investigate the impact of different language models on typing rate and accuracy.