Motion Laboratory

In this study, the motion of nunchaku action was expressed by 3DCG as the motion of a tool moving along with the motion of a human body. The basic motion of nunchaku action was measured using an optical motion capture system, and an attempt was made to reproduce the action of vertically swinging nunchaku with 3DCG.The motions of the human body and the nunchaku were measured at the same time, and the body's motions were then edited and output as data in the form of a skeleton.Three markers were attached to the tips of the two handles of the nunchaku, and the position and angle of the club part of the nunchaku were were calculated from the position coordinates of these markers using the inner product and outer product.The motion of the human body and the nunchaku were integrated into a single development environment and expressed as CG animation.Six people evaluated the created CG animation and answered a questionnaire on their understanding of and interest in the nunchaku performance. From the results, although the demonstration video gave an incomplete impression without the nunchaku's chain part, the effects of intriguing the participants and deepening the their understanding of their motion were achieved as expected.

In this research, we develop a system to design and try on Japanese kimono by operating 3D characters in VR space with an HMD and controllers. In this system, users can design Japanese kimono by arranging the pattern of materials prepared in front of the design area of the VR space by scaling it with the controllers. This system allows users to draw a line with a pen, or change the background color of the kimono. The design is reflected in the Japanese kimono worn by a 3D character in real time, and users can intuitively preview the results while virtually trying it on. The friction and movement of the sleeves are reproduced by physics-based simulation, and the sleeves flutter according to the motion of the user's hands. 3D human characters were prepared with three different sleeve lengths to evaluate the differences in design and how the sleeves flutter. From a third-person perspective, the motion of the 3D character is reproduced, and users can objectively appreciate the state of the moving sleeves. In order to verify the operability and usefulness of this system, an evaluation experiment was conducted. As a result, about 80% of the respondents gave positive answers, confirming that this system could effectively stimulate interest in kimono.

In 2020, there is a problem that multiple people cannot be in the same space due to the influence of the new coronavirus infection. In this research, we proposed a dance choreography creation system that allows users who are far away to collaborate remotely. In this system, a PC connected to a network is used for remote collaborative work, and choreography is created by continuously playing back any number of selected steps. In addition, in this system, by synchronizing between users, multiple users can work while watching the same video. In addition, in this system, by synchronizing between users, multiple users can work while watching the same video. Users can use text chat when creating choreography, discuss the steps to be selected between users, and enter the steps selected by the user to look back on who selected what step. In this system, in order to reduce the lag when synchronizing, the user not only sends the result of the movement of the 3D model, but also sends the execution command from input to playback. We asked nine students to experience this system and conducted an evaluation experiment. As a result, 80% of the subjects answered that this system can support and assist in the creation of dance choreography, confirming that this system is useful for collaborative work between users who are far away.

Since cultural properties are valuable objects, it is difficult to actually pick them up and browse their details or check their interiors. In addition, it is hard to form a vivid impression just by looking at them in museum exhibitions. We propose an interface that can roll out a developing view of a cultural property from a cylindrical object that adopts VR for the purpose of arousing interest and promoting understanding of such cultural properties. Using this interface, the developing view from the cylindrical object can be rolled out by pulling two controllers apart, and it can be stored by moving the controllers together. In this way, rotating the cylindrical object varies the length of the rolled-out image, and thus the pulled-out developed view and the cylindrical object are synchronized. To make it easier for users to operate the system intuitively, the developing view is rolled out from the position closest to the user on the lateral surface of the cylindrical object. Using the proposed interface, we created VR content based on the Shari casket as an example implementation. In order to evaluate the usefulness of this interface, an experiment was conducted using nine students as evaluators, and about 60% of the students answered that their interest and understanding had deepened, thus confirming the potential of this interface.

In this study, the author developed a system that allows users to enjoy and experience the thinking of Chunibyo expressed in CG with the aim of promoting understanding of the youth language of Chunibyo.This system uses Kinect to recognize five gestures that are said to be acted by people with so-called Chuunibyo.With this system, the 3D object that visualizes the thoughts of Chunibyo is displayed and animated with effect corresponding to the recognition result.In the gesture of "slashing" that swings the arm vigorously, this system calculates the speed from the moved distance of a hand position, and recognizes shooking the arm when the speed of the hand is above a certain level and slower than the previous frame.Then, this system displays an animation effect of slashing in the direction of the arm swung.The author made it possible to always recognize any gesture by making the recognition conditions of the five gestures dissimilar.In order to evaluate the usefulness of this system, I asked eight subjects to actually experience the system. As a result, the author received manypositive evaluations.Therefore, the author found that he was able to achieve the purpose of promoting understanding of the meaning of Chunibyo while having fun. However, it was also found that further ingenuity was required for the accuracy of gesture recognition.

Folklore recounts a festival in which Ryukyu dance was performed on ships in Ryutan pond during the Ryukyu Kingdom era, but it is difficult to actually reproduce this festival in the current age.The purpose of this study is to reproduce this historical Ryukyu dance with CG and to arouse interest in the festival. This study proposes a dance performance authoring system for Ryukyu dance using an AR (augmented reality) headset. In this system, Ryutan pond, ships, and dancers are created by CG and superimposed on a floor in the real world, and the user operates ships and creates the motions of ships and dancers. In order to create a dance intuitively, this system allows users to arrange ship and dancer models by pointing to a position with a pointer of the controller and to create ship animation by specifying keyframes. The Ryukyu dance assigned to a dancer was performed by a professional dancer and acquired by motion capture with the aim of making it similar to the dance as it was originally performed in the era of the festival.Eight students actually experienced the system and evaluated it. As a result, 90% of the students answered that they were able to arrange and move ships as intended. Consequently, this system confirmed that users could create simple performances and move ships.

In this study, I develop a system for joining body-part motions using reinforcement learning with the aim of supporting the creation of choreography for contemporary dance.The system generates short choreographic motions and displays animations of them. The generated motions are learned based on feature values using reinforcement learning. The motion features in this study are defined as the average features per frame for one motion. The feature values for a single motion are calculated by summing the rotational amount of each joint with a weight for each body part of all joints. In order to verify that learning with different weights generates motions with different characteristics, an evaluation of impressions was conducted by having subjects compare generated motions. As a result, it was found that the larger the weight given, the greater the immobility.

In this study, the author aims to automatically generate dance choreography that is useful for choreographers and to use random forest for analyzing the factors that make various combinations of choreography useful. As training data, time-series information on the start and end of the choreography combination was added to data of the choreography combination generated by a body-part motion synthesis system (BMSS), and then the choreographer's five-level evaluation data were applied to training. The data were classified into three types of data according to the data evaluation value: for use as it is, whether it can be used in the dance performance, and whether it is attractive. A trained model of each pattern was generated using a random forest. Useful choreography combinations were analyzed from the obtained importance and evaluation, types of choreography combinations, synthesis timing, and the number of syntheses. From the analysis results, the author confirmed that the evaluation by choreography was high depending on the specific movement and timing. In addition, the author clearly identified movements that tend to be usable in dance performance as well as movements that tend to be attractive when synthesized.