Enhanced EEG-based cognitive workload detection using RADWT and machine learning

Area of Science:

• Neuroscience
• Cognitive Science
• Machine Learning

Background:

• Understanding cognitive workload is crucial for optimizing learning and human cognitive process analysis.
• Accurate estimation of cognitive workload has significant applications in adaptive learning, brain-computer interfaces, and cognitive monitoring.

Purpose of the Study:

• To investigate different cognitive workload levels and propose a classification approach using Rational-Dilation Wavelet Transform (RADWT).
• To evaluate various machine learning and feature selection techniques for optimal classification accuracy in cognitive workload estimation.

Main Methods:

• Utilized Rational-Dilation Wavelet Transform (RADWT) for analyzing electroencephalogram (EEG) signal sub-bands, focusing on temporal and spectral dynamics.
• Employed machine learning classifiers, including Linear Support Vector Machine (LSVM), and minimum Redundancy Maximum Relevance (mRMR) for feature selection.
• Identified the frontal brain region and alpha/theta rhythm sub-bands as most relevant for differentiating cognitive workload levels.

Main Results:

• Achieved high classification accuracies: 96.6% (0-back vs. 3-back), 94.9% (0-back vs. 2-back), 92.3% (2-back vs. 3-back), and 81.7% (three-class scenario).
• Demonstrated the effectiveness of integrating RADWT with LSVM and mRMR for precise cognitive workload estimation.
• Highlighted the frontal region and specific EEG rhythms (alpha, theta) as key indicators of cognitive workload.

Conclusions:

• The proposed RADWT-based machine learning approach is a valid and effective method for estimating cognitive workload.
• Findings provide insights into the neural mechanisms underlying cognitive workload.
• The study lays a foundation for developing advanced adaptive systems, brain-computer interfaces, and cognitive monitoring tools.