Resumen

HERNANDEZ GARCIA, Ruber; GARCIA REYES, Edel; RAMOS COZAR, Julián  y  GUIL MATA, Nicolás. Features representation models for human actions classification in video: state of art. Rev cuba cienc informat [online]. 2014, vol.8, n.4, pp. 21-51. ISSN 2227-1899.

Human actions classification in video is a very active investigation area in computer vision. The objective of this research area is to classify automatically human actions from the frames that make up a video sequence, using pattern recognition techniques. The performance of pattern recognition methods is heavily dependent on the choice of data representation on which they are applied. For this reason, this paper focuses on the analysis of the state of the art concerning the representation models of visual information for human actions classification. This paper aims to critically analyze the different approaches reported and their theoretical aspects. Finally, the study concluded that the application of features selection techniques, the use of relational models and obtaining representation based on visual n-grams shown as interesting alternatives to incorporate as part of representation models for human actions classification.

Palabras clave : features representation; features selection; human actions classification; visual vocabularies.

        · resumen en Español     · texto en Español     · Español ( pdf )