Enhancing classification accuracy of fNIRS-BCI using features acquired from vector-based phase analysis.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Brain-computer interfaces (BCIs) enable communication and control through brain activity.
• Functional near-infrared spectroscopy (fNIRS) is a non-invasive neuroimaging technique measuring hemodynamic responses.
• Improving classification accuracy in fNIRS-BCIs is crucial for practical applications.

Purpose of the Study:

• To present a novel feature extraction methodology for fNIRS-based BCIs.
• To enhance classification accuracy for two-class and three-class fNIRS-BCI tasks.
• To validate the effectiveness of the proposed features.

Main Methods:

• Novel features were extracted using a vector-based phase analysis method.
• Four novel features derived from oxygenated and de-oxygenated hemoglobin changes were calculated.
• fNIRS signals from motor cortex during finger tapping tasks were analyzed using linear discriminant analysis.

Main Results:

• The combination of four novel features achieved significantly higher average classification accuracies: 98.7% for two-class and 85.4% for three-class BCIs.
• These accuracies substantially outperformed conventional features, which yielded 68.7% and 53.6% respectively.
• Validation on an open-access database confirmed improved classification accuracies for both two-class and three-class fNIRS-BCIs.

Conclusions:

• The proposed vector-based phase analysis method offers a significant advancement in fNIRS-BCI feature extraction.
• This novel approach leads to substantial improvements in classification accuracy for fNIRS-BCIs.
• The findings represent a step forward in enhancing the performance of state-of-the-art fNIRS-BCIs.