Segmenting motion capture data using a qualitative analysis

Abstract

Many interactive 3D games utilize motion capture for both character animation and user input. These applications require short, meaningful sequences of data. Manually producing these segments of motion capture data is a laborious, time-consuming process that is impractical for real-time applications. We present a method to automatically produce semantic segmentations of general motion capture data by examining the qualitative properties that are intrinsic to all motions, using Laban Movement Analysis (LMA). LMA provides a good compromise between high-level semantic features, which are difficult to extract for general motions, and lowlevel kinematic features, which often yield unsophisticated segmentations. Our method finds motion sequences which exhibit high output similarity from a collection of neural networks trained with temporal variance. We show that segmentations produced using LMA features are more similar to manual segmentations, both at the frame and the segment level, than several other automatic segmentation methods. University of Pennsylvania AbstractMany interactive 3D games utilize motion capture for both character animation and user input. These applications require short, meaningful sequences of data. Manually producing these segments of motion capture data is a laborious, time-consuming process that is impractical for real-time applications. We present a method to automatically produce semantic segmentations of general motion capture data by examining the qualitative properties that are intrinsic to all motions, using Laban Movement Analysis (LMA). LMA provides a good compromise between high-level semantic features, which are difficult to extract for general motions, and lowlevel kinematic features, which often yield unsophisticated segmentations. Our method finds motion sequences which exhibit high output similarity from a collection of neural networks trained with temporal variance. We show that segmentations produced using LMA features are more similar to manual segmentations, both at the frame and the segment level, than several other automatic segmentation methods.

Deep motifs and motion signatures

“…Subsequences are identified through frequency analysis and compared via dynamic time warping in order to cluster similar sequences. More recently, Bouchard and Badler [2015] segment motions semantically by examining their LMA-inspired qualitative properties.…”

seg1d: A Python package for Automated segmentation of one-dimensional (1D) data

“…With feature-rich data such as multiple marker trajectories from motion capture, the reduction of meaningful features is important for segmentation performance (Bouchard & Badler, 2015). The use of marker trajectory and ground reaction force, without computing kinematics, has been shown to be sufficient in movement segmentation tasks (Lin, Bonnet, Joukov, Venture, & Kulic, 2016).…”

Continuous body emotion recognition system during theater performances

“…The ability of unsupervised learning made it appropriate for experimentation. Many scientists used neural networks in emotion recognition , and the work of Bouchard et al on body motion segmentation was the first attempt to justify their classification criteria. As the neural network shows their efficiency and adaptability for facial and body expression recognition, we believe it is the appropriate tool for body emotion recognition.…”