A 240-target VEP-based BCI system employing narrow-band random sequences

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computer Science

Background:

• Brain-computer interfaces (BCI) are crucial for assistive technologies.
• High information transfer rates (ITR) and large instruction sets are key challenges in current BCI systems.
• Code-modulated visual evoked potential (c-VEP) BCIs offer a promising avenue for improved performance.

Purpose of the Study:

• To develop a novel c-VEP BCI system capable of handling an extensive instruction set.
• To maintain high performance metrics, including information transfer rates (ITR).
• To address the challenge of achieving high ITR with a large number of targets in BCI.

Main Methods:

• A novel c-VEP BCI system utilizing narrow-band random sequences as visual stimuli.
• A convolutional neural network (CNN)-based EEG2Code decoding algorithm for stimulus sequence prediction.
• Both offline (sequential paradigm) and online (cued spelling task) experiments were conducted.

Main Results:

• Offline experiments achieved an average accuracy of 87.66% and a simulated ITR of 260.14 bits/min.
• Online experiments demonstrated an accuracy of 76.27% and an ITR of 213.80 bits/min.
• The system supported one of the largest known instruction sets for VEP-based BCIs.

Conclusions:

• The developed c-VEP BCI system represents an advancement in the field.
• The system exhibits robust performance, offering high accuracy and ITR.
• This work has the potential for more practical and efficient brain-computer interface applications.