Enhancing SSVEP Identification With Less Individual Calibration Data Using Periodically Repeated Component Analysis.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Steady-state visually evoked potentials (SSVEP) identification methods often struggle with small calibration datasets.
• Existing methods like task-related component analysis (TRCA) rely on inter-trial components, which can be limited with sparse data.

Purpose of the Study:

• To introduce a novel method, Periodically Repeated Component Analysis (PRCA), for robust SSVEP identification.
• To enhance SSVEP identification accuracy and efficiency, particularly when calibration data is scarce.

Main Methods:

• PRCA constructs spatial filters to maximize reproducibility across periods and creates synthetic SSVEP templates from periodically repeated components (PRCs).
• PRCs were integrated into improved TRCA variants.
• Performance was evaluated on 16-target and 40-target datasets, and in an online experiment.

Main Results:

• PRCA-based methods achieved over 95% and 90% accuracy with only 1-second data and a single calibration trial per frequency.
• High information transfer rates (ITR) were recorded: up to 198.8±57.3 bits/min and 191.2±48.1 bits/min.
• Online experiments yielded 94.00 ± 7.35% accuracy and 139.73±21.04 bits/min ITR with 0.5-second calibration data.

Conclusions:

• PRCA-based methods demonstrate significant performance improvements with reduced calibration data.
• These methods are effective and robust for SSVEP identification, showing potential for practical SSVEP-BCI applications.