Pattern Recognition of Cognitive Load Using EEG and ECG Signals

Area of Science:

• Neuroscience
• Educational Psychology
• Biomedical Engineering

Background:

• Effective learning hinges on matching cognitive load with working memory capacity.
• Cognitive effort during learning elicits measurable physiological responses.
• Automated recognition of cognitive load patterns using physiological data is valuable.

Purpose of the Study:

• To develop and evaluate automatic pattern recognition models for cognitive load states using physiological measures.
• To identify key physiological features indicative of cognitive load and its matching with working memory.
• To compare the performance of different classifiers for cognitive load state recognition.

Main Methods:

• Extraction of 33 physiological features to quantify autonomic and central nervous system activity.
• Selection of a critical feature subset using sequential backward selection and particle swarm optimization.
• Construction and comparison of pattern recognition models using decision tree and support vector machine classifiers.

Main Results:

• Decision tree achieved 96.3% accuracy in distinguishing cognitive load from baseline states.
• Support vector machine achieved 97.2% accuracy in differentiating cognitive load mismatching from matching states.
• Key distinguishing features included active state of mind and autonomic nervous system activity.

Conclusions:

• Physiological measures can reliably distinguish between cognitive load and baseline states.
• Cognitive load matching and mismatching states are identifiable through specific physiological patterns.
• This approach supports the development of adaptive learning systems based on real-time cognitive load assessment.