Pseudo-online detection and classification for upper-limb movements.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computer Science

Background:

• Decoding human movement intentions is crucial for advanced brain-computer interfaces (BCIs).
• Previous BCIs often struggle with real-time accuracy and robustness in distinguishing movement types.

Purpose of the Study:

• To analyze movement detection and classification for decoding upper-limb volitions in a pseudo-online setting.
• To develop and evaluate an ensemble processing pipeline for improved BCI performance.

Main Methods:

• Nine healthy subjects performed self-initiated upper-limb movements.
• Evaluated three classifiers (SVM, EEGNET, Riemannian geometry SVM) across different frequency bands for movement detection.
• Developed an ensemble model using majority voting and an adaptive boosted Riemannian geometry model for classification.

Main Results:

• The ensemble model achieved a 79.6% true positive rate with low false positives (3.1 per minute) and minimal latency (75.3 ms).
• Movement classification accurately differentiated contralateral movements with approximately 67% accuracy.

Conclusions:

• The proposed ensemble method offers a robust approach for brain-computer interface design.
• The pseudo-online testing procedure validates the system's potential for practical BCI applications.