Interactive Techniques For High Level Spacetime Editing Of Human Locomotion

Interactive Techniques for High-level Spacetime Editing of Human Locomotion

In recent years, computer animated characters have become commonplace, often appearing extremely lifelike in movies, television, and video games. Unfortunately, animating digital characters remains difficult and time-consuming, pursued mostly by dedicated professionals, because conventional animation methods require careful control and coordination of many different low-level parameters which affect the believability of a character's pose and timing. In this thesis, we present techniques to allow high-level control of motions for animated human characters, by interactively editing existing locomotive motions. These techniques prioritize interactivity by allowing the user to manipulate more general or high-level aspects of the motion, such as a motion path. These techniques also have low data and computation requirements, as they operate on a single input motion without expensive optimization or simulation, allowing for instant feedback. The foundation of this work is a novel path-based editing algorithm based on simple and well-established biomechanical and physical principles of motion. After user manipulation of a motion path, the poses of the character along the path are automatically adjusted to match the new path, and an automatic timewarping procedure adjusts the timing of the new motion depending on its context, in order to remain consistent with the original even after large changes. Our two performance techniques utilize familiar input methods to control the path-based editing algorithm in different ways. First, based on a study exploring how users can express motion using their hands, our "finger walking" technique allows control of motion editing by mimicking full-body motion on a touch-sensitive tabletop. A further study was conducted to test automatic identification of the motion type of the performance, and to evaluate user satisfaction with the generated motions. The second technique uses handheld manipulation of a mobile device, such as a tablet or smartphone, to mimic a motion through space. Using only the simple and standard motion sensors in these devices, we developed two gestural methods for editing motion, with both discrete, one-off gestures and continuous "steering wheel" style control of an ongoing motion. Overall, the techniques presented in this thesis demonstrate that fast and interactive editing of human locomotion is possible using meaningful high-level controls, without requiring large amounts of additional motion data or computation.

Computer Animation and Simulation ’96

The 14 papers in this volume vividly demonstrate the current state of research in real-time animation. Half of the papers are dedicated to algorithm allowing the real-time animation of complex articulated structure in particular (humans, legged robots, plants) and of dynamic scenes in general. The proposed approaches cover from motion capture to motion reusability which are essential issues for high-end applications as 3D games, virtual reality, etc. Other topics treated are motion management for fast design of realistic movements, 2D and 3D deformations, and various optimization techniques for simulation (adaptive mass-spring refinement, huge particule systems).

Automated Methods for Data-driven Synthesis of Realistic and Controllable Human Motion