Motion Laboratory

A dancing group requires a formation, quite unlike a solo dance, and thus the arrangement on the stage is important. We have developed a system to simulate the formations of a dancing group. The simulation is executed by specifying the number of dancers, the formation pattern, and dancing motion with a game pad. Typical formations used for classical ballet pieces were analyzed, and six formation patterns were obtained. Each dancer's position is calculated by the number of dancers and the selected forms. The system arranges dancers on a virtual stage and simulates dance animations by using motion-capture data obtained using a professional dancer.

Our goal is to produce live stage performances featuring interactions between actors and CG images projected on the stage. We have developed a system that allows actors to interactively control projected three-dimensional CG animation while performing Noh dance. This system can change the CG animation on a screen in real time from the values obtained by a three-axis accelerometer attached to a Noh actor's foot. This system detects two kinds of basic Noh motions from the actor's jerk calculated by the acceleration. The actor can add CG characters to the projected images and change their motions by making the corresponding Noh motions. This system was used for the "Utoh" hell scene in a Noh-style stage performance of "A Spider's Thread."

In dance training, the difference between professional dancers and intermediate-level dancers is difficult to perceive and explain by looking at their dancing figures only. In this research, we focus on one female dancer. The differences in classical ballet skills are quantitatively analyzed and compared using motion data captured at different ages of the dancer. The target motion is a basic step called an arabesque, and the data were obtained in August 2008 (age 15) and in August 2010 (age 17). Twenty-seven physical features are defined and calculated. Then they are analyzed by principal component analysis to examine whether a special trend appears. As a result, small differences in the captured motion are found between the different dates.

The purpose of our study is to distinguish between dances quantitatively. It is assumed that the human body is composed of 25 ball-joints. The location data of the 25 ball-joints is evaluated from the motion data which is obtained by motion capture system. Then the acceleration data are evaluated by smoothing second derivative of the location data. And the angle data are evaluated by using inner product of body part vectors. Eigen vectors are obtained by applying principal component analysis for the acceleration data and the angle data. For three class of dance (Noh, classical ballet, contemporary dance), the difference between the eigenvectors of each dance are examined. And the expansion coefficients are examined by evaluating.

• Tsuyoshi Matsuoka, Asako Soga, Kazuhiro Fujita, A Retrieval System for Ballet Steps using Three-dimensional Motion Data, Proc. of the 5th International Conference on Intelligent Information Hiding and Multimedia Signal Processing, pp.1144-1147 (Kyoto, Japan), Sep. 2009