Harmonic Component Analysis: A novel training-free and asynchronous BCI classification method.

Area of Science:

• Neuroscience and Biomedical Engineering
• Assistive Technology Development

Background:

• Existing brain-computer interfaces (BCI) for locked-in syndrome patients often lack reliability and require extensive user training.
• There is a need for training-free BCI systems capable of asynchronous and online control for assistive robotic technologies.

Purpose of the Study:

• To investigate a novel training-free BCI classifier, Harmonic Component Analysis (HCA), for asynchronous control of assistive technologies.
• To evaluate the performance of HCA against Canonical Correlation Analysis (CCA) for steady-state visually evoked potentials (SSVEP) detection.

Main Methods:

• Developed and proposed the Harmonic Component Analysis (HCA), a training-free classifier for harmonic-characteristic signals like SSVEP.
• Compared HCA with a three-component Canonical Correlation Analysis (CCA) using an offline dataset from 10 healthy participants with 16 SSVEP targets.

Main Results:

• HCA demonstrated superior performance to CCA, with up to 74% lower computational cost.
• For asynchronous control, HCA achieved 85% detection accuracy (1.6s activation) versus 77% for CCA (1.7s activation).
• HCA showed a higher true positive rate (65% vs 59%) with a lower false positive rate (0.59% vs 0.27%) for continuous activation.

Conclusions:

• Harmonic Component Analysis (HCA) is a suitable SSVEP classifier for asynchronous classification without requiring calibration or training sessions.
• The HCA offers a promising, efficient, and reliable solution for brain-computer interfaces (BCI) in assistive technologies.