Toward a Real Time View-invariant 3D Action Recognition - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2016

Toward a Real Time View-invariant 3D Action Recognition

Résumé

In this paper we propose a novel human action recognition method, robust to viewpoint variation, which combines skeleton-and depth-based action recognition approaches. For this matter, we first build several base classifiers, to independently predict the action performed by a subject. Then, two efficient combination strategies , that take into account skeleton accuracy and human body orientation, are proposed. The first is based on fuzzy switcher where the second uses a combination between fuzzy switcher and aggregation. Moreover, we introduce a new algorithm for the estimation of human body orientation. To perform the test we have created a new Multiview 3D Action public dataset with three viewpoint angles (30°,0°,-30°). The experimental results show that an efficient combination strategy of base classifiers improves the accuracy and the computational efficiency for human action recognition.
Fichier principal
Vignette du fichier
conf_Mounir.pdf (2.34 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-01332468 , version 1 (15-06-2016)

Identifiants

Citer

Mounir Hammouche, Enjie Ghorbel, Anthony Fleury, Sébastien Ambellouis. Toward a Real Time View-invariant 3D Action Recognition. International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP), Feb 2016, Roma, Italy. ⟨10.5220/0005843607450754⟩. ⟨hal-01332468⟩
109 Consultations
153 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More