MoRTELaban: a Neurosymbolic Framework for Motion Representation and Analysis based on Labanotation and Laban Movement Analysis

Abstract

Human motion cannot be fully modeled by subsymbolic representations.While these extract precise hidden patterns in motion data, they are often task-specific and lack a semantic understatement of motion.Symbolic systems that mirror human cognition and explicit expressive processes are necessary for richer motion synthesis and analysis, enabling physical reasoning and expert knowledge encoding.In this work, we propose a neurosymbolic framework that combines Labanotation and Laban Movement Analysis (LMA), originally developed for dance, to represent and analyze human motion symbolically.We expand the existing LabanEditor to support full-body annotation and integrate it with AMASS, Mediapipe, and Kinect inputs through a SMPL-based format.Our system supports automatic annotation for the local functional and expressive aspects of motion, and enables bidirectional conversion between symbols and motion.While still a work in progress, this framework lays the groundwork for explainable, expressive motion modeling that can support human-robot interaction, motion preservation, and psychomotor learning systems.