MOTUS – Automated Analysis and Prediction of Human Movement Qualities

Computational methods for analyzing and predicting expressive, affective, and social aspects of human movement are crucial for developing technologies that enhance quality of life and support clinical decision-making. This multidisciplinary, curiosity-driven project aims to advance movement analysis through automated detection and tracking of expressive qualities, with a focus on the Origin of Movement (OoM)—the body part perceived as initiating motion. OoM detection offers valuable insights for applications in healthcare and beyond.

The project explores novel machine learning approaches (e.g., hierarchical clustering, matrix completion, and graph neural networks) for OoM analysis, with validation in a key healthcare case study: psychiatrist-patient interactions. Measuring attunement—the responsiveness between individuals—is critical for predicting clinical outcomes, particularly in psychiatry, where the Therapeutic Relationship (TR) influences treatment adherence. However, attunement is complex, spanning multiple spatio-temporal scales. To address this, we apply our computational methods to three validated tasks assessing behavioral synchronization, coordination, and cooperation, offering new insights into attunement and laying the groundwork for AI-assisted mental healthcare.