Exemplar-based Parametric Hidden Markov Models for Recognition and Synthesis of Movements

Dennis Herzog, Volker Krüger, Daniel Grest

Publikation: Bidrag til bog/antologi/rapport/konference proceedingKonferenceartikel i proceedingForskningpeer review

Abstrakt

A common problem in movement recognition is the recognition of movements of a particular type. E.g. pointing movements are of a particular type but differ in terms of the pointing direction. Arm movements with the goal of reaching out and grasping an object are of a particular type but differ with the location of the involved object.
In this paper, we present an exemplar-based parametric hidden Markov model (PHMM) that is able to recognize and synthesize movements of a particular type. The PHMM is based on exemplar movements that have to be ``demonstrated'' to the system. Recognition and synthesis are carried out through locally linear interpolation of the exemplar movements. Experiments are performed with pointing and grasping movements. Synthesis is done based on the object position as parameterization. In case of the recognition, the coordinates of the grasped or pointed at object are recovered. Our experiments show the flexibility of our exemplar-based PHMMs in terms of the amount
of training data and its robustness in terms of noisy observation data.
OriginalsprogEngelsk
TitelProceedings of Vision, Modeling, and Visualization 2007
Antal sider261
ForlagDie Deutsche Bibliothek
Publikationsdato2007
Sider253
ISBN (Trykt)978-3-940739-00-1
StatusUdgivet - 2007
BegivenhedVision, Modeling, and Visualization 2007 (VMV) - Saarbrücken, Tyskland
Varighed: 7 nov. 20079 nov. 2007

Konference

KonferenceVision, Modeling, and Visualization 2007 (VMV)
LandTyskland
BySaarbrücken
Periode07/11/200709/11/2007

    Fingerprint

Citationsformater

Herzog, D., Krüger, V., & Grest, D. (2007). Exemplar-based Parametric Hidden Markov Models for Recognition and Synthesis of Movements. I Proceedings of Vision, Modeling, and Visualization 2007 (s. 253). Die Deutsche Bibliothek.