Examining Four Experimental Paradigms for EEG-Based Sleep Quality Evaluation with Domain Adaptation

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Dry electrode electroencephalography (EEG) offers objective sleep quality assessment.
• Lack of a standard experimental paradigm impedes EEG sleep evaluation development.

Purpose of the Study:

• To evaluate four typical experimental paradigms for EEG-based sleep quality assessment.
• To develop a new EEG dataset using a dry-electrode headset.
• To address individual variations using domain adaptation (DA) methods.

Main Methods:

• Experimental examination of four distinct EEG acquisition paradigms.
• Development of a novel EEG dataset with dry electrodes.
• Application of domain adaptation techniques, including Domain Adversarial Neural Network (DANN) and Transfer Component Analysis (TCA).

Main Results:

• A relaxing paradigm demonstrated superior performance over attention-concentration paradigms.
• The relaxing paradigm achieved an average accuracy of 76.01% for sleep quality evaluation.
• DANN improved accuracy by 18.69% compared to TCA in domain adaptation.

Conclusions:

• A relaxing experimental setup is optimal for EEG-based sleep quality evaluation.
• Domain adaptation methods, particularly DANN, are effective in mitigating inter-subject variability.
• The developed dry-electrode EEG dataset and findings contribute to advancing objective sleep assessment techniques.