![](bg1.png)
A COMPREHENSIVE ANALYSIS OF HUMAN MOTION CAPTURE
DATA FOR ACTION RECOGNITION
Valsamis Ntouskos, Panagiotis Papadakis and Fiora Pirri
ALCOR, Vision, Perception and Cognitive Robotics Laboratory, Department of Computer and System Sciences,
University of Rome ”La Sapienza”, Rome, Italy
Keywords:
Motion Analysis, Motion Capture, Action Recognition.
Abstract:
In this paper, we present an analysis of human motion that can assist the recognition of human actions irrespec-
tive of the selection of particular features. We begin with an analysis on the entire set of preclassified motions
in order to derive the generic characteristics of articulated human motion and complement the analysis by
a more detailed inter-class analysis. The statistical analysis concerns features that describe the significance-
contribution of the human joints in performing an action. Furthermore, we adopt a hierarchical analysis on the
human body itself in the study of different actions, by grouping joints that share common characteristics. We
present our experiments on standard databases for human motion capture data as well as a new commercial
dataset with additional classes of human motion and highlight certain interesting results.
1 INTRODUCTION
Research interest has been largely stimulated by the
analysis of human motion as it constitutes a key com-
ponent in a plurality of disciplines. Common applica-
tions range from human action recognition, human-
machine interaction, skill learning and smart surveil-
lance to applications within the entertainment indus-
try such as character animation, computer games and
film production.
As manifested by earlier research (Mihai, 1999),
(Kovar and Gleicher, 2004), (Thomas et al., 2006),
(Poppe, 2010) the study of human motion is not a new
research field, however, most of the focus has so far
been directed towards 2D image-based human motion
representations. In contrast, with the evolution of mo-
tion capture hardware-software and recently with the
advent of affordable depth acquisition devices such
as the Kinect (Microsoft, 2010), part of research is
shifting towards 3D (spatial) representations of hu-
man motion extracted from a hierarchical represen-
tation of the human pose, i.e. a skeletal structure.
Motion capture data (MOCAP) contain informa-
tion of the human pose as recorded during the exe-
cution of an action that is organized into a collection
of 3D points-joints together with the spatial position
or rotation of the corresponding coordinate frames.
These points comprise a skeletal representation of the
human pose wherein the motion of the points has been
acquired either directly from body sensors (optical or
magnetic) or tracked along the action sequence.
Previous studies on the characteristics of articu-
lated human motion such as (Fod et al., 2002), (Pullen
and Bregler, 2002), (Jenkins and Mataric, 2003),
(Barbi
ˇ
c et al., 2004), (Okan, 2006) have been con-
ducted within various contexts, namely, 3D animation
compression, motion synthesis and motion indexing-
classification or segmentation. However, experiments
are usually performed on relatively small collections
of human motions with relatively limited variability
in the classes of motion.
In this work, we provide a comprehensive study
of human motion, through a statistical analysis of a
diverse set of human action categories using MOCAP
databases. Our experiments are performed on stan-
dard databases (CMU, 2003), (Muller et al., 2007)
and further extended on a commercial database (mo-
capdata.com, 2011), altogether highlighting impor-
tant characteristics of articulated human motion re-
lated to the correlation of human motion and hierar-
chy of the human kinematic chain in terms of body
part contribution. In particular, we perform a statisti-
cal analysis on various abstraction levels of the joint
representation of the human skeleton across different
datasets and investigate the variance within the in-
herent dimensionality of simple and complex actions.
The motivation of our study is to provide a deeper in-
sight into the characteristics of articulated human mo-
647
Ntouskos V., Papadakis P. and Pirri F..
A COMPREHENSIVE ANALYSIS OF HUMAN MOTION CAPTURE DATA FOR ACTION RECOGNITION.
DOI: 10.5220/0003868806470652
In Proceedings of the International Conference on Computer Vision Theory and Applications (VISAPP-2012), pages 647-652
ISBN: 978-989-8565-03-7
Copyright
c
2012 SCITEPRESS (Science and Technology Publications, Lda.)