Motion Similarity Analysis and Evaluation of Motion Capture Data

  • Author(s) / Creator(s)
  • Technical report TR05-11. Motion similarity analysis is a critical stage for the successful reuse of motion capture data. Some previous works use one or multiple motion features, such as the difference between joint positions, joint angles, joint velocities and accelerations to capture the similarity information between two frames of different or the same motion streams. In this paper, two typical motion similarity approaches are reviewed and related problems are identified. To address the problems, a novel motion similarity method is proposed and two different features are used to measure the similarity between two motion frames: the curvature of space curve and the difference between joint relative positions. Additionally, we also propose a new general criterion to evaluate and to compare the performance of different motion similarity techniques. The experimental results demonstrate that our new motion similarity method can automatically generate visually acceptable results efficiently and that the other methods can be improved by using our proposed evaluation criterion to select the required attribute weights appropriately. | TRID-ID TR05-11

  • Date created
  • Subjects / Keywords
  • Type of Item
  • DOI
  • License
    Attribution 3.0 International