A high-speed hybrid brain-computer interface with more than 200 targets.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Human-Computer Interaction

Background:

• Brain-computer interfaces (BCIs) are advancing, with target/command count being crucial for decoding user intentions.
• Existing BCI systems have not surpassed 200 targets, limiting their practical applications.
• High-speed BCI systems are needed to enhance user experience and expand application scenarios.

Purpose of the Study:

• To develop the first high-speed BCI system capable of decoding over 200 targets.
• To integrate multiple electroencephalography (EEG) features for enhanced BCI performance.
• To evaluate the system's accuracy and information transfer rate in spelling tasks.

Main Methods:

• Developed a hybrid BCI system using P300, motion visual evoked potential (mVEP), and steady-state visual evoked potential (SSVEP) features.
• Employed a time-frequency division multiple access strategy to encode 216 targets.
• Utilized task-discriminant component analysis and linear discriminant analysis for feature decoding.
• Conducted offline and online spelling experiments with human subjects.

Main Results:

• Offline analysis confirmed prominent P300 and mVEP in central, parietal, and occipital regions, with SSVEP strongest in the occipital region.
• Online experiments demonstrated an average accuracy of 85.37% (cued-guided) and 86.00% (free-spelling) for 216-target classification.
• Achieved average information transfer rates (ITR) of 302.83 bits/min (cued-guided) and 204.47 bits/min (free-spelling), with a peak ITR of 367.83 bits/min.

Conclusions:

• The developed BCI system is the first to achieve over 200 targets at high speed.
• The hybrid feature approach and advanced decoding methods significantly enhance BCI performance.
• This breakthrough holds promise for broadening the application scope of BCI technology.