Phase synchronization for the recognition of mental tasks in a brain-computer interface

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Surface electroencephalogram (EEG)-based brain-computer interfaces (BCIs) are crucial for communication in motor-disabled populations.
• Current BCIs often rely on spectral estimates and autoregressive models for feature extraction.
• Investigating novel features is essential for enhancing BCI performance.

Purpose of the Study:

• To evaluate the utility of EEG signal synchronization for classifying mental tasks.
• To compare the performance of phase locking value (PLV) and spectral coherence features against traditional power spectral density estimates.
• To determine the optimal combination of features for improved BCI accuracy.

Main Methods:

• Analysis of five 60-minute EEG recordings from three subjects performing three distinct mental tasks.
• Investigation of features derived from phase locking value (PLV) and spectral coherence.
• Comparison with classification rates from alpha, beta1, beta2, and 8-30-Hz power spectral densities.
• Offline analysis without artifact removal or rejection.

Main Results:

• Significant differences were observed between PLV and mean spectral coherence measures.
• Classification accuracies of up to 62% were achieved using synchronization measures alone.
• Combining phase synchronization measures with alpha power spectral density estimates yielded the best classification results.
• Phase synchronization provides valuable information for classifying spontaneous EEG during mental tasks.

Conclusions:

• Phase synchronization is a relevant feature for improving mental task classification in EEG-based BCIs.
• The integration of synchronization measures with spectral power enhances BCI performance.
• This study highlights the potential of phase synchronization for future BCI development in assistive communication.