Classification of code-modulated visual evoked potentials using adaptive modified covariance beamformer and EEG...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Code-modulated Visual Evoked Potentials (c-VEP) enable high Information Transfer Rate (ITR) Brain-Computer Interfaces (BCIs).
• Spatiotemporal Beamforming (STB) is a common approach for decoding EEG in c-VEP BCIs.
• Conventional STB methods can struggle with robust covariance matrix estimation in short stimulation times.

Purpose of the Study:

• To develop a novel STB-based technique for improved gazed target detection in c-VEP BCIs.
• To enhance the performance of STB by enabling robust covariance matrix estimation with limited data (short stimulation times).
• To evaluate the effectiveness of user parameter-free methods for covariance matrix estimation.

Main Methods:

• A novel STB-based technique was proposed for c-VEP BCI target detection.
• User parameter-free methods (convex combination, general linear combination, and modified versions) were employed for robust covariance matrix estimation.
• The proposed methods were assessed using a stimulus presentation rate of 120 Hz and minimal repetitions of m-sequences.

Main Results:

• The proposed STB methods significantly improved classification accuracy by an average of 20% compared to conventional STB at the shortest stimulation times.
• An average Information Transfer Rate (ITR) of 157.07 bits/min was achieved using only two repetitions of m-sequences.
• The novel technique demonstrated superior performance across all tested stimulation times.

Conclusions:

• The developed STB-based technique offers a significant performance improvement over conventional methods for c-VEP BCIs.
• Robust covariance matrix estimation with limited data is achievable using the proposed user parameter-free methods.
• This advancement holds promise for more efficient and accurate c-VEP BCI systems.