Evaluation of EEG features in decoding individual finger movements from one hand

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Noninvasive brain-computer interfaces (BCIs) are advancing rapidly.
• A key limitation is the scarcity of movement-related features for control signals.
• Previous studies explored electrocorticography (ECoG) and electroencephalography (EEG) for finger decoding.

Purpose of the Study:

• To evaluate multiple movement-related features for discriminating individual finger movements using noninvasive EEG.
• To identify novel EEG features for enhanced BCI control.

Main Methods:

• Noninvasive EEG data was collected during a finger discrimination task.
• Spectral Principal Component Analysis (PCA) was applied to extract spectral features.
• EEG temporal data and spectral power changes in alpha and beta bands were also analyzed.

Main Results:

• A broadband feature in EEG was identified for individual finger discrimination, previously only seen in ECoG.
• Spectral features from PCA achieved an average decoding accuracy of 45.2% (P < 0.05).
• This accuracy was significantly higher than chance and other tested features like temporal EEG data and band-specific spectral power.

Conclusions:

• Noninvasive EEG can decode individual finger movements, offering a richer set of control features.
• The identified broadband EEG feature and spectral PCA features hold significant promise for advancing noninvasive BCI applications.
• This research facilitates the development of more complex and sophisticated noninvasive BCI systems.