Johannes Mezger

On the representation, learning and transfer of spatio-temporal movement characteristics

In this paper we present a learning-based approach for the modeling of complex movement sequences. Based on the method of Spatio-Temporal Morphable Models (STMMs) we derive a hierarchical algorithm that, in a first step, identifies automatically movement elements in movement sequences based on a coarse spatio-temporal description, and in a second step models these movement primitives by approximation through linear combinations of learned example movement trajectories. We describe the different steps of the algorithm and show how it can be applied for modeling and synthesis of complex sequences of human movements that contain movement elements with a variable style. The proposed method is demonstrated on different applications of movement representation relevant for imitation learning of movement styles in humanoid robotics.

Interactive physically-based shape editing

We present an alternative approach to standard geometric shape editing using physically-based simulation. With our technique, the user can deform complex objects in real-time. The basis of our method is formed by a fast and accurate finite element implementation of an elasto-plastic material model, specifically designed for interactive shape manipulation. Using quadratic shape functions, we reduce approximation errors inherent to methods based on linear finite elements. The physical simulation uses a volume mesh comprised of quadratic tetrahedra, which are constructed from a coarser approximation of the detailed surface. In order to guarantee stability and real-time frame rates during the simulation, we cast the elasto-plastic problem into a linear formulation. For this purpose, we present a corotational formulation for quadratic finite elements. We demonstrate the versatility of our approach in interactive manipulation sessions and show that our animation system can be coupled with further physics-based animations like, e.g. fluids and cloth, in a bi-directional way.

Estimation of Skill Levels in Sports Based on Hierarchical Spatio-Temporal Correspondences

We present a learning-based method for the estimation of skill levels from sequences of complex movements in sports. Our method is based on a hierarchical algorithm for computing spatio-temporal correspondence between sequences of complex body movements. The algorithm establishes correspondence at two levels: whole action sequences and individual movement elements. Using Spatio-Temporal Morphable Models we represent individual movement elements by linear combinations of learned example patterns. The coefficients of these linear combinations define features that can be efficiently exploited for estimating continuous style parameters of human movements. We demonstrate by comparison with expert ratings that our method efficiently estimates the skill level from the individual techniques in a ”karate kata”.

CGForum 2008 Cover Image

After the silver-coloured feet had been fixed on the ground, the deformation was modelled during a short interactive session. A dedicated plasticity model facilitates realistic, intuitive modelling while volumetric effects are appropriately incorporated. The surface mesh consists of 250.000 vertices (original mesh courtesy of Stanford Computer Graphics Laboratory), which are transformed smoothly and interactively by the quadratic shape functions of the underlying tetrahedral mesh.