Assessment of visual fatigue in SSVEP-based brain-computer interface: a comprehensive study.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• User fatigue significantly degrades brain-computer interface (BCI) performance.
• Existing electroencephalographic (EEG) fatigue detection methods yield inconsistent results, especially in steady-state visually evoked potential (SSVEP) BCIs.
• A need exists for robust fatigue detection to improve BCI reliability.

Purpose of the Study:

• To identify factors contributing to inconsistent fatigue detection results in the literature.
• To investigate fatigue detection in SSVEP-based BCIs using extended experimental durations and varied stimulation.
• To propose novel and reliable EEG features for fatigue detection.

Main Methods:

• Conducted an experiment on an SSVEP-BCI system designed to induce user fatigue.
• Analyzed EEG signals from O1, Oz, and O2 channels.
• Calculated traditional EEG features (rhythm powers, SNR) and novel features (spectral features, Lempel-Ziv complexity).

Main Results:

• Observed a shift from high-frequency to low-frequency EEG rhythms with fatigue.
• Identified 'relative power' of EEG rhythms, specific frequency ratios (e.g., θ/β), spectral features (central frequency, asymmetry), and Lempel-Ziv complexity as promising indicators.
• These features demonstrated a behavior consistent with the observed frequency shift.

Conclusions:

• The proposed EEG features, including relative power, frequency ratios, spectral characteristics, and Lempel-Ziv complexity, offer a more trustworthy approach to fatigue detection.
• These features can be utilized to develop a more reliable fatigue index for SSVEP-BCI systems.
• Addressing fatigue is essential for enhancing the overall performance and usability of BCI systems.