ML model for fatigue prediction in brain-computer interface applications through SSVEP analysis

Abstract Fatigue assessment is crucial in brain-computer interface (BCI) applications, as it helps ensure the reliability, safety, and effectiveness of the system. Objective methods for fatigue assessment using electroencephalogram (EEG) analysis provide valuable insights into the user's cognitive state and can help optimize the function of a BCI system. Within our research, a new fatigue assessment approach utilizing fractal dimensions and spectral analysis was presented to assess the subject’s fatigue level in a designed steady-state visual evoked potential (SSVEP)-based BCI experiment. To elicit SSVEPs, visual stimuli were delivered using nine flickering cues with frequencies of 6, 8, 10, 12, 15, 18, 20, 25, and 30 Hz to 26 healthy volunteers during EEG recording. Naïve Bayes classifier using fractal dimension attributes succeeded in classifying fatigue and alert states with a high accuracy of 97.31% at the stimulation frequency of 15 Hz. Specifically, the experimental outcomes showed that the Petrosian fractal dimension, with a high accuracy of 97.59%, can be a potential biomarker for fatigue prediction in SSVEP-based BCIs. Hence, we acknowledge the suitability of the Petrosian fractal dimension for objectively evaluating fatigue when utilizing an SSVEP-based BCI.