Common spatial pattern and wavelet decomposition for motor imagery EEG- fTCD brain-computer interface.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Hybrid brain-computer interfaces (BCIs) combine multiple modalities to improve performance over single-modal BCIs.
• Previous work introduced a hybrid BCI using electroencephalography (EEG) and functional transcranial Doppler ultrasonography (fTCD) for motor imagery (MI) tasks.

Purpose of the Study:

• To enhance the accuracy and information transfer rate (ITR) of hybrid BCIs.
• To investigate a novel probabilistic fusion method for combining EEG and fTCD signals.
• To compare the performance of the proposed hybrid BCI with existing systems.

Main Methods:

• Utilized multi-scale analysis and common spatial pattern (CSP) algorithm for feature extraction from EEG and fTCD data.
• Proposed a probabilistic fusion approach using a Bayesian framework to combine EEG and fTCD evidence.
• Evaluated the system on motor imagery tasks (right MI vs. baseline, left MI vs. baseline, right MI vs. left MI).

Main Results:

• Achieved average accuracies of 93.85%, 93.71%, and 100% for the respective MI tasks.
• Attained average information transfer rates (ITRs) of 19.89, 26.55, and 40.83 bits/min.
• Demonstrated superior performance compared to previous studies using the same hybrid BCI and outperformed EEG-fNIRS systems.

Conclusions:

• The proposed hybrid EEG-fTCD BCI system offers significant improvements in accuracy and speed.
• The probabilistic fusion method enhances the effectiveness of combining multimodal brain signals.
• This system presents a more accurate and faster alternative to existing EEG-fNIRS based BCIs.