close

The Department of Computer Science at the University of Cyprus cordially invites you to the Colloquium entitled:

Self-similarity Analysis to Auto-correct Motion Capture Data

 

Speaker: Dr. Andreas Aristeidou
Affiliation: University of Cyprus, Cyprus
Category: Colloquium
Location: Room 148, Faculty of Pure and Applied Sciences (FST-01), 1 University Avenue, 2109 Nicosia, Cyprus (directions)
Date: Wednesday, November 23, 2016
Time: 11:30-12:30 EET
Host: Yiorgos Chrysanthou (yiorgos-AT-cs.ucy.ac.cy)
URL: https://www.cs.ucy.ac.cy/colloquium/index.php#cs.ucy.2016.aristeidou

Abstract:
In this talk, we propose a novel method for self-similarity analysis of raw motion capture sequences. The recent advances in motion capture (MoCap) technology allow the acquisition of highly dynamic movements, such as dances and gymnastics that comprise of a large repertoire of arbitrary and complex poses. However, despite technological advances, there are still many instances where occlusions and noise can lead to missing and erroneous data. This problem is more pronounced in complex or constrained movements (e.g., for modern dances) or when more than one subject is captured (e.g., dancing in pairs in flamenco or salsa). Wrong and missing marker positioning can create abnormal motion and outliers in a number of joints in the reconstructed motion. Our analysis allows auto-detecting outliers and abnormal joint rotations as well as correcting them automatically. The key idea relies on the premise that the expected motion data has high-degree of self-similarity. The presented method is analogous to patch-based self-similarity techniques used in images and video. Here, instead of patches we use motion words, which consist of short-periodic-sequences of all joints transformations. What makes our approach particularly interesting is that motion words, in contrast to text and image, are continuous and can vary in length. Moreover, the large number of possible movement combinations, the irregularity of human actions, and the variability in style of different people makes the analysis of the motion and its correction a challenging problem.

Short Bio:
Dr. Andreas Aristidou is a Post-Doc researcher at the Efi Arazi School of Computer Science, The Interdisciplinary Center Herzliya (Israel), and the Graphics & Virtual Reality Lab, Department of Computer Science, University of Cyprus. He had been a Cambridge European Trust fellow, at the Department of Engineering, University of Cambridge, where he obtained his PhD. Andreas has a BSc in Informatics and Telecommunications from the National and Kapodistrian University of Athens, and he is an honor MSc graduate of Kings College London. He has been awarded a number of prestigious awards, including the ΔΙΔΑΚΤΩΡ fellowship for young researchers from the Cyprus Research Promotion Foundation, the Office of Naval Research Visiting Award, and the DARIAH-EU Theme 2015 in Open Humanities. Andreas collaborates with PhaseSpace Inc., a leading company that offers motion capture solutions for motion tracking and positioning, while he establishes a motion capture laboratory at the Graphics & Virtual Reality Lab, University of Cyprus. His research interests lie on motion analysis and classification, motion synthesis, and involve motion capture and Inverse Kinematics.

  Web: https://www.cs.ucy.ac.cy/colloquium/
  Mailing List: https://listserv.cs.ucy.ac.cy/mailman/listinfo/cs-colloquium
  RSS: https://www.cs.ucy.ac.cy/colloquium/rss.xml
  Calendar: https://www.cs.ucy.ac.cy/colloquium/schedule/cs.ucy.2016.aristeidou.ics

Sponsor: The CS Colloquium Series is supported by a generous donation from Microsoft