Towards Adaptive ERP-Based BCIs: EEG and fNIRS Guided Flashing Strategies to Account for Cognitive Fatigue

This pilot study presents a hybrid EEG–fNIRS framework to

jointly assess speller performance and cognitive fatigue.

Speller paradigms often face a trade-off: increasing stimulus

flash repetitions improves ERP signal-to-noise ratio (SNR)

but prolongs task duration, while fewer flashes reduce

efficiency under fatigue. To address this limitation, EEG and

fNIRS signals were recorded simultaneously during speller

and cognitive task, arithmetic problem-solving blocks.

Behavioral findings showed a decline in performance of the

cognitive task, with fewer correct answers and longer

response times across blocks, suggesting increased fatigue.

ERP analyses revealed overall strong target responses but

with attenuated amplitudes over time, especially in parietal

and occipital channels. Classification confirmed these

patterns: target vs. non-target discrimination reached 98.6%

accuracy with three EEG channels, cognitive block

classification achieved 92.4% using EEG, and 87.4% using

fNIRS. These results demonstrate the feasibility of hybrid

EEG–fNIRS systems for monitoring user states and provide

a foundation for adaptive speller BCIs that dynamically

adjust stimulus repetitions.