Inferring Mental States via Linear and Non-Linear Body Movement Dynamics: A Pilot Study.

Area of Science:

• Human-computer interaction
• Occupational health
• Biomechanical analysis

Background:

• Occupational tasks in sports, healthcare, military, and transportation involve stress, workload, and uncertainty.
• Coordinated whole-body movements may reflect these transient mental states.
• Wearable sensors and motion capture can quantify movement dynamics to classify mental states impacting performance and human-machine interaction.

Purpose of the Study:

• To test the feasibility of using movement dynamics to classify mental states.
• To identify preliminary movement features linked to stress, workload, and uncertainty.
• To explore the potential of movement-based sensing for applied settings and human-machine interface development.

Main Methods:

• A pilot study (N=10) involved participants performing a facial emotion recognition task under variable stress, workload, and uncertainty.
• Movement trajectories were analyzed holistically and by phase, extracting up to 3844 features across multiple domains.
• A 10-fold cross-validation procedure with feature selection (T-Rex Selector) and machine learning (Logistic Regression) was employed.

Main Results:

• Movement-based models achieved mean ROC AUC scores up to 0.76 for stress classification.
• Highest sensitivity for stress detection was observed during the full movement trajectory and middle (raise) phases.
• Classification of workload and uncertainty states was less successful.

Conclusions:

• Movement-based sensing shows potential for inferring stress states in applied occupational settings.
• Preliminary findings support the use of movement dynamics for understanding mental states.
• Further research can inform the development of advanced human-machine interfaces.