Comparison of Trajectories and Quaternions of Folk Dance Movements Using Dynamic Time Warping

Abstract

Analysis and comparison of human motion are very important for games and virtual reality. In the case of dances, teacher's or professional performance is considered as the ground truth. To provide feedback to the learner, learner's movements are compared to the teacher's movements. Recording human motion using optical motion capture system is often considered as gold standard. In this paper, the problem of getting ground truth data has been analyzed. Professional dancers, male and female, were recorded performing folk dances, alone and with a partner using an optical motion capture system. Two cases for dancing in pair, one person was wearing the suit with passive markers, and both were wearing suits. Three-dimensional marker trajectories and quaternions were compared using dynamic time warping and multidimensional dynamic time warping. Initial results show that dances performed in pair are the most similar and could be used in the applications for learning purposes.