An SSVEP-based BCI with 112 targets using frequency spatial multiplexing

Area of Science:

• Neuroscience
• Computer Science
• Biomedical Engineering

Background:

• Brain-computer interface (BCI) systems face challenges in achieving high target resolution, limiting their practical applications.
• Steady-state visual evoked potential (SSVEP) based BCIs offer potential for numerous targets but are hindered by stimulus competition.
• Improving target resolution is crucial for advancing BCI capabilities and meeting application demands.

Purpose of the Study:

• To overcome the limitations of stimulus competition in SSVEP-based BCIs.
• To propose and validate a frequency spatial multiplexing method for increasing target resolution.
• To enhance the performance of BCI systems by improving the number of accessible commands.

Main Methods:

• Developed a frequency spatial multiplexing paradigm by arranging flicker stimuli as 2x2 matrices in a tiled interface.
• Designed and tested three distinct interface layouts using the proposed paradigm.
• Implemented a graph neural network to differentiate targets of the same frequency based on EEG response patterns.

Main Results:

• Experimental validation with eleven subjects demonstrated the effectiveness of the proposed method.
• Average offline classification accuracies reached up to 91.38% across three paradigms.
• Achieved high information transfer rates (ITR), with the highest being 53.96 bits/min.

Conclusions:

• The frequency spatial multiplexing method successfully increases target resolution in SSVEP BCIs by leveraging stimulus spatial relationships.
• The developed graph neural network effectively distinguishes targets, improving classification accuracy.
• This approach provides a foundation for further advancements in SSVEP detection efficiency and BCI performance.