LEMON
Overview
Title: | Towards an episodic situation understanding – Learning and real-time catEgorization of complex MOtioN trajectories exemplified by human action sequences |
---|---|
Duration: | 01.10.2009 - 31.10.2013 |
Funding: | Honda Research Institute Europe Offenbach |
Project Partner: | Honda Research Institute Europe Offenbach |
Project Manager: | Prof. Dr. Horst-Michael Groß, Dr.-Ing. Klaus Debes |
Staff: | Dipl.-Inf. Christian Vollmer |
Description
The focus of this project is the development of methods for visual human action recognition on a mobile robot platform operating in a domestic environment. In this project a human action is to be understood as a small-scale movement of one ore more limbs on a relative short timescale, e.g. throwing a ball or picking up something.
To be able to perform motion analysis the robot has to observe, predict, and classify movements of the observed person. In this project movements will be represented entirely by motion trajectories of the limbs. Thus the work is focussed on real-time prediction and classification of two or three dimensional time series.
An important emerging task of robot systems in the domain of elderly care is to maintain the cognitive abilities of the person cared for. One of many ways to achieve cognitive stimulation is to do physical exercises. A robot system equipped with the ability to classify and predict motion can help a person by guiding it throught such exercises and giving feedback about the quality of the executed exercises.
Publications
Papers available for download on the lab's publication page
Journals
Vollmer, Ch., Hellbach, S., Eggert, J., Gross, H.-M.
Sparse Coding of Human Motion Trajectories with Non-negative Matrix Factorization
to appear: Neurocomputing (2013), Elsevier, doi: http://dx.doi.org/10.1016/j.neucom.2012.12.054http://dx.doi.org/10.1016/j.neucom.2012.12.054
Conferenes
2012
Vollmer, Ch., Eggert, J., Gross, H.-M.
Generating Motion Trajectories by Sparse Activation of Learned Motion Primitives.
in: Proc. 22. Int. Conf. on Artificial Neural Networks (ICANN 2012), Lausanne, Switzerland, Part I, LNCS 7552, pp. 637-644, Springer 2012
Vollmer, Ch., Eggert, J., Gross, H.-M.
Modeling Human Motion Trajectories by Sparse Activation of Motion Primitives Learned from Unpartitioned Data.
in: Proc. 35th German Conference on Artificial Intelligence (KI 2012), Saarbrücken, LNCS 7526, pp. 168-179, Springer 2012
2011
Hellbach, S., Vollmer, Ch., Eggert, J., Gross, H.-M.
Learning Motion Primitives using Spatio-Temporal NMF.
in Proc. of the Workshop New Challenges in Neural Computation 2011, Machine Learning Reports 05/2011, pp. 5-8