Parametric HMMs for Movement Recognition and Synthesis

Dennis Herzog, Volker Krüger

Research output: Contribution to book/anthology/report/conference proceedingArticle in proceedingResearchpeer-review

Abstract

A common problem in human movement recognition is the recognition of movements of a particular type (semantic). E.g., grasping movements have a particular semantic (grasping) but the actual movements usually have very different appearances due to, e.g., different grasping directions. In this paper, we develop an exemplar-based parametric hidden Markov model (PHMM) that allows to represent movements of a particular type. Since we use model interpolation to reduce the necessary amount of training data, we had to develop a method to setup local models in a synchronized way.
In our experiments we combine our PHMM approach with a 3D body tracker. Experiments are performed with pointing and grasping movements parameterized by their target positions at a table-top. A systematical evaluation of synthesis and recognition shows the use of our approach. In case of recognition, our approach is able to recover the movement type, and, e.g., the object position a human is pointing at. Our experiments show the flexibility of the PHMMs in terms of the amount of training data and its robustness in terms of noisy observation data. In addition, we compare our PHMM to an other kind of PHMM, which has been introduced by Wilson and Bobick.
Original languageEnglish
Title of host publicationProceedings of IEEE's NTAV/SPA
Number of pages6
Publication date2009
Pages9-15
Publication statusPublished - 2009
EventNTAV / SPA - Poznan, Poland
Duration: 25 Sep 200827 Sep 2008

Conference

ConferenceNTAV / SPA
CountryPoland
CityPoznan
Period25/09/200827/09/2008

Keywords

  • action recognition
  • action representation
  • parametric hidden Markov models
  • computer vision
  • robotics

Fingerprint Dive into the research topics of 'Parametric HMMs for Movement Recognition and Synthesis'. Together they form a unique fingerprint.

  • Projects

    Perception, Action & Cognition through Learning of Object-Action Complexes

    Dillmann, R., Eklundh, J. O., Wörgötter, F., Krüger, V. K. C., Krüger, N., Ude, A., Torras, C., Hommel, B. & Steedman, M.

    <ingen navn>

    01/02/200631/01/2010

    Project: Research

  • Cite this

    Herzog, D., & Krüger, V. (2009). Parametric HMMs for Movement Recognition and Synthesis. In Proceedings of IEEE's NTAV/SPA (pp. 9-15)