Improve the Classification Efficiency of High-Frequency Phase-Tagged SSVEP by a Recursive Bayesian-Based Approach

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Steady-state visual evoked potential (SSVEP)-based brain-computer interfaces (BCIs) offer reliable human-computer interaction.
• Phase-tagged SSVEP (p-SSVEP) extends target availability, particularly for high-frequency BCIs.
• Current high-frequency p-SSVEP recognition efficiency is limited by trade-offs between accuracy and information transfer rate.

Purpose of the Study:

• To introduce and evaluate a novel recursive Bayesian-based approach for improving high-frequency p-SSVEP classification efficiency.
• To address the low recognition efficiency and computational time constraints in high-frequency p-SSVEP BCIs.
• To enhance the accuracy and practical bit rate of SSVEP-based BCIs.

Main Methods:

• Developed a recursive Bayesian-based classification approach incorporating dynamic prior probabilities.
• Implemented the approach for processing SSVEP stimuli at high frequencies (20 Hz and 30 Hz) with six phases each.
• Compared the recursive Bayesian method against three other classification approaches using identical data segment lengths.

Main Results:

• The recursive Bayesian-based approach achieved the highest classification accuracy and practical bit rate compared to other methods.
• Mean accuracy reached 89.7% with a 37.8 bits/min practical bit rate at 20 Hz.
• Mean accuracy was 89.0% with a 36.5 bits/min practical bit rate at 30 Hz, demonstrating reduced individual differences.

Conclusions:

• The recursive Bayesian-based approach significantly improves classification efficiency for high-frequency p-SSVEP BCIs.
• This method offers a promising solution for achieving higher accuracy and bit rates while minimizing inter-subject variability.
• The approach enhances the practical usability of high-frequency SSVEP BCIs.