Estimating Systemic Cognitive States from a Mixture of Physiological and Brain Signals

Area of Science:

• Human-Computer Interaction
• Cognitive Science
• Machine Learning

Background:

• Human-machine teams require artificial systems to detect human cognitive states for optimal performance.
• Physiological and neurophysiological signals (e.g., heart rate, EEG) are linked to cognitive states like workload and distraction.
• The sufficiency of these signals alone for accurate state inference across individuals is an open research question.

Purpose of the Study:

• To introduce an experimental and machine learning framework to investigate the inference of human cognitive states.
• To evaluate the effectiveness of machine learning techniques using physiological and neurophysiological data.
• To establish a baseline for future improvements in cognitive state classification.

Main Methods:

• Developed a multitasking interactive experimental setting to collect multimodal data.
• Utilized physiological (heart rate, skin conductance) and neurophysiological (EEG, fNIRS) measurements.
• Applied standard machine learning techniques to classify cognitive states such as cognitive load and mind wandering.

Main Results:

• Standard machine learning methods achieved modest classification success using only physiological and neurophysiological signals.
• Results indicate the complexity of inferring cognitive states across subjects based solely on biosignals.
• The findings provide a baseline for future research, particularly for methods incorporating contextual data.

Conclusions:

• Inferring human cognitive states from physiological and neurophysiological signals alone presents significant challenges.
• Contextual information (task and environmental states) is likely essential for improving classification accuracy.
• This study establishes a foundation for developing more robust human-aware artificial systems.