Discriminative codewaves: a symbolic dynamics approach to SSVEP recognition for asynchronous BCI

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Steady-state visual evoked potential (SSVEP) is a key brain-computer interface (BCI) technology.
• Current SSVEP methods often require user training and recording of driving signals for accurate gaze direction prediction.
• High performance in SSVEP-based BCIs typically relies on multichannel signal analysis guided by these driving signals.

Purpose of the Study:

• To introduce an efficient method for identifying attended stimuli in SSVEP-based BCIs.
• To eliminate the need for registering driving signals during SSVEP analysis.
• To enhance the practicality and user-friendliness of SSVEP communication pathways.

Main Methods:

• Brain activity, specifically regional brain response, is modeled as a dynamical trajectory towards attractors.
• Discriminative vector quantization is employed for a condensed description of single-trial responses.
• A classification scheme using templates and confidence intervals from a small training set disentangles different brainwave trajectories.

Main Results:

• The introduced approach demonstrates favorable comparison with established alternatives in terms of information transfer rate.
• Experiments using two distinct datasets validated the effectiveness of the novel method.
• The method achieves high performance in identifying attended stimuli without driving signal registration.

Conclusions:

• The novel approach offers an efficient way to identify attended stimuli in SSVEP-based BCIs.
• It utilizes a unique description of brainwaves based on semi-supervised learning and single sensor traces.
• The method holds significant potential for developing self-paced and more accessible BCIs.