Predicting Supine Infant Center-of-Pressure from Pose Estimation using Physics and Data Driven Modeling
Infants with neuromotor impairments often struggle more to maintain proper postural balance compared to their typically developing counterparts. Studies have shown promising results in quantifying infant impairment by observing changes in their Center of Pressure (COP). However, the force plates necessary to capture COP are often not readily accessible due to price and portability. Although COP is a good summary of infant motion and balance, little is known about how specific motions contribute to infant COP. In response, this project aims to (1.) make COP easier to obtain by predicting supine infant COP from human pose information gathered with cameras, and (2.) quantify the relationship between infant motion and COP, gaining insight into the nature of infant postural control and impairment. To accomplish both these goals, we have derived a generalized physics-informed model of the infant’s dynamics that calculates COP from pose-based kinematics. We anticipate that this physics model will include some errors due to generalizations and assumptions made. To minimize model error and further improve prediction ability, we will augment our results using data-driven models working in concert with our physics-informed model.
