Searching 3d Motion Patterns of Vietnamese Traditional Dances

Abstract

Vietnam has many traditional dances suchas Xoan singing, “tuồng” or “chèo”. They all urgentlyneed to be preserved in digital formats, especially in3D motion capture format for dances. In digitalformats, they bring many values such as the ability toautomatically classify and search for content ofdances' movement. In this paper, we propose asystem for 3D movement search of Cheo dance 'spostures and gestures. The system applies slidingwindow technique, Dynamic Time Warpingalgorithm and a novel feature selection methodnamed CheoAngle. Results show that the proposedsystem reach good scores in several metrics. We alsocompare CheoAngle with other feature selectionmethods for 3D movement and show that CheoAnglegive the best results