Dictionary of gray-level 3D patches for action recognition

Stefen Chan Wai Tim (1) , Michèle Rombaut (1) , Denis Pellerin (1)

1 GIPSA-AGPIG - GIPSA - Architecture, Géométrie, Perception, Images, Gestes

Sparse dictionary-based methods have received a lot of focus in the recent years. In this paper, we are interested in action recognition in videos using a 3D representation based on the use of a dictionary. A 3D volume can be seen as a set of gray-level 3D patches which are 2D patches taken in successive frames in order to capture a motion pattern. The goal is to recognize human actions within these 3D volumes, whose 3D patches are described with the dictionary atoms. A motion signature is computed by building a histogram based on the use of the atoms of the dictionary. Paired with a SVM, we show that these signatures can be exploited in the context of action recognition. This method was tested on the KTH database with good results.

Domaines: Traitement du signal et de l'image [eess.SP]; Traitement du signal et de l'image [eess.SP]

Stefen Chan Wai Tim, Michèle Rombaut, Denis Pellerin. Dictionary of gray-level 3D patches for action recognition. MLSP 2014 - IEEE 24th International Workshop on Machine Learning for Signal Processing, Sep 2014, Reims, France. ⟨hal-01058339⟩