Automatic annotation of traditional dance data using motion features

This study investigates the accuracy of human dance motion capture and classification from selected Malaysian dances for Malaysian Dance Annotation (MDA). Dance motion classification is a new scope to motion classification. Recent studies focus on basic movements classification where motion is not very complex, such as walking or waving hand. In this paper, an attempt has been made to classify complex motion such as a dance. This work is motivated by the need to develop an automated tool to annotate dance videos to build traditional dance video management and retrieval system. We evaluate our system on a dataset of different types of Malaysian dances, collected from Youtube. Despite the complex movements in dance, the proposed solution requires less human input effort and gets a suitable accuracy for complex dance motion annotation.