On the Detection of Video's Ethnic Vietnamese Thai Dance Movements

The problem addressed in this paper is the one of classifying Vietnamese dances' videos. In particular, we focus on an automatic detection of movements in the Ethnic Vietnamese Thai dances (ETVD). We first propose an ontology-based description of ETVD movements in terms of main movements' steps. We then associate with each movement step a profile containing typical features that characterize a movement step. The automatic detection of ETVD movements is based on a correlation method that matches movements' steps profiles with concepts present in frames of dances' videos. The last part of the paper contain experimental studies that show the good classification rate of our ETVD movement detection method.