Correlation-based channel selection for cognitive workload assessment and classification using EEG signals

Area of Science:

• Neuroscience
• Cognitive Science
• Biomedical Engineering

Background:

• Cognitive workload, the mental effort in tasks, impacts daily decisions and efficiency.
• Electroencephalography (EEG) signals offer a reliable, non-invasive method for measuring cognitive workload.
• Accurate cognitive workload estimation is crucial for optimizing performance and minimizing errors.

Purpose of the Study:

• To classify cognitive workload levels using EEG signals.
• To reduce computational complexity for real-time applications through channel selection.
• To evaluate the effectiveness of time-frequency decomposition techniques for workload assessment.

Main Methods:

• Channel selection using Pearson Correlation Coefficient.
• Time-frequency decomposition via Maximal Overlap Discrete Wavelet Transform (MODWT) and Empirical Mode Decomposition (EMD).
• Feature extraction and selection using Improved Distance Evaluation, followed by classification with SVM, K-Nearest Neighbors, and Decision Tree.

Main Results:

• Frontal EEG channels were identified as critical for cognitive workload assessment.
• A combination of MODWT and EMD with the Support Vector Machine (SVM) classifier achieved the highest accuracy.
• Optimal channel selection and time-frequency methods significantly improved classification accuracy while reducing system complexity.

Conclusions:

• Frontal EEG channels are highly informative for cognitive workload estimation.
• Hybrid time-frequency decomposition (MODWT-EMD) coupled with SVM offers a robust approach for workload classification.
• This methodology enhances accuracy and efficiency for real-time cognitive workload monitoring.