Pattern Classification of Instantaneous Cognitive Task-load Through GMM Clustering, Laplacian Eigenmap, and Ensemble...

Area of Science:

• Human-Computer Interaction
• Neuroscience
• Machine Learning

Background:

• Accidents in human-machine systems can result from unmanaged cognitive task-load (CTL).
• Neurophysiological data and machine learning offer objective methods for recognizing discrete CTL levels during operation.

Purpose of the Study:

• To design subject-specific, multi-class CTL classifiers.
• To uncover the relationship between operator performance and neurophysiological features.
• To improve safety in human-machine collaborative systems.

Main Methods:

• Collected psychophysiological data during human-machine control tasks.
• Defined 4-5 CTL classes using Gaussian mixture models and performance variables.
• Extracted EEG features with Laplacian eigenmap and used heart rate.
• Ensemble classifier created by aggregating support vector machines via majority voting.

Main Results:

• The proposed method successfully derived target CTL classes.
• Identified a low-dimensional set of optimal EEG features for individual operators.
• Demonstrated the capability of the ensemble classifier in recognizing CTL classes.

Conclusions:

• The developed subject-specific CTL classifier effectively recognizes cognitive task-load levels.
• The approach provides a promising tool for enhancing safety in human-machine systems.
• Feature selection and ensemble methods are effective for CTL classification.