Motion Laboratory

Since Maki-e fountain pens are degraded by light and humidity, they are unsuitable for permanent exhibitions. Some fine patterns are also difficult to see with the naked eye. In recent years, VR technology and the digital archiving of museum materials have advanced. In this study, we developed a VR-appreciating system for Maki-e fountain pens to support museum exhibitions. In our system, the recreated 3DCG fountain pens of Maki-e and a rollout image, which is a continuous picture, can be simultaneously previewed in the VR world. Our system allows users to preview Maki-e fountain pens on VR and extract rollout images by pulling apart a Maki-e fountain pen with both controllers. Users can hold Maki-e fountain pens in the VR world and change the pen’s scale by pressing a button on the controller. We evaluated our experiment with visitors who experienced our system. Questionnaire results showed that about 70% believed that the appreciation method using VR was appropriate. About 40% thought that the interface was difficult to understand. Therefore, this study confirmed that although the current interface has room for improvement, the concept of viewing fountain pens in VR is appropriate.

In recent years, with the popularity of idols and artist groups, interest in formation dances, which form a group of people and attract dances as a group has increased. In order to create a formation of multiple dancers and simulate their traveling with 3DCG, I proposed a system that has some functions to control the traveling of dance groups. In this system, the position of multiple CG dancers and their destinations are set on the virtual stage, and human motion and traveling of group dance are displayed in 3DCG animation. In order to control the body orientation of each dancer, the system calculates the traveling orientation from the position of each dancer to each destination, displays the dancer's front orientation every frame by interpolating the orientation. In addition, a function has been provided to replace the positions of multiple dancers clockwise or counterclockwise from the current formation in order to achieve the rotation traveling that is often used in formation dance. In order to confirm the usefulness of the system, I conducted an evaluation experiment with 8 subjects. As a result, I obtained a result that this system is useful because it can create dance formations and preview the traveling of multiple dancers.

The purpose of this study is to generate dance choreographies that can be useful for Choreographers using machine learning. In this study, I analyzed what factors are highly evaluated for short choreography combinations generated by the Body-part Motion Synthesis System that is automatically create combinations of body-part movement. As the learning data, I used choreography combination data and its evaluation data by a professional choreographer. I used random forests for machine learning and generated a learning model. Importance rating of each motion that affects for evaluation by the choreography was calculated from generated learning model. I analyzed the tendency of evaluated choreography combinations that has high importance rating. As a result of analysis, I found motion that have a positive impact.

I analyzed the use of a body-part motion synthesis system to explore ways to use it for creating dance performances by choreographers. From the log data of the choreography sequences created by a choreographer, I tabulated the types and numbers of basic motions and body-part motions used in the motion synthesis and compared the 3DCG video with the demonstration videos of works created based on the choreography sequences in the first and second dance productions of the same choreographer. My analysis found that the choreographer often chose two body-part motions that were synthesized to the basic motion. In a comparison between 3DCG and demonstration videos, the proportion of identical motions decreased and the proportion of motion changes increased.

For practicing kendo, both an opponent and guidance from an experienced teacher are required. I developed a system that allows users to experience kendo using virtual reality technologies to learn its basic actions and techniques. The user wears a head-mounted display (HMD), holds a controller in hand, and wears trackers on both feet. The user faces opponent in the virtual reality world, operates a bamboo sword that moves in conjunction with the controller, and strikes the head, arm, and trunk of opponent. CG animation of the basic action is displayed next to the user in the VR space. During a series of input actions of the users to the opponent, swing motions are judged using head and hand heights obtained from the HMD and the controller. The feet arrangement, which is critical in kendo, is also judged using the position of both feet obtained from the trackers. 27 subjects used this system and evaluated its effectiveness. About 90% provided positive answers to questions about the CG animation’s operability and example actions. About 80% provided positive answers to questions related to the position of the input actions. I also received comments that focused on areas of the system that must be improved: the accuracy of its judgment of the input action and simplifying the user interface display.

This study detects abnormal walking that is difficult to observe visually. I developed a system that visualize the feature values of human motion data with 3DCG. It calculates waist direction, knee angle, and ankle angle as the feature values for each frame from the 3-dimensional coordinate data of markers captured by an optical motion capture system, which allows users to simultaneously preview human motion and the visualized feature values by 3DCG animation. To easily see the required features , in each walking phase the system can change the displayed feature values by time. Since Initial Contact is the moment that your foot reaches the ground, left knee and right ankle are visualized as planes, and the waist direction is visualized as lines．Initial Swing is the period from when the toe leaves the ground until it crosses the opposite leg, the system visualizes right knee as planes and the waist directions. Iexperimentally evaluated the usefulness of my system’s concept with six people and got such positive comments as it was easy to understand because the important parts were visualized in the walking phases.

In this study, I compared the practice swing motions using motion data to clarify the features of unique practice swing. Human motions of normal swing, reverse possession swing, and large shares swing of one player were measured 10 times using an optical motion capture system. The opening angle of the armpit, the rotation angle of the waist, and the speed of the bat tip were calculated from the captured 3D data as feature values. The rotation angle of the waist was calculated as the angle between waist vector facing front at the pose and each waist vector in swinging. As the result of comparing the calculated feature values, the rotation angle of the waist in practice swing of large shares is larger than that of normal swing. I found that practice swing of large shares is that the waist can be rotated greatly and the player can swing stable. The angle of the aside on practice swing of reverse possession is smaller than that of normal swing. Therefore, I found that practice swing of reverse possession is that can be swing with the armpit tightened.