Identifier
etd-07102015-222844
Degree
Master of Science in Engineering Science (MSES)
Department
Engineering Science (Interdepartmental Program)
Document Type
Thesis
Abstract
Text to animation is the conversion of textual instructions to animations following and depicting those instructions. Text to animation is important to fields such as crime scene investigation, military special operations and storytelling where text is utilized. It is also useful for physical therapy where a doctor can factor in disability parameters in the 3D models and determine the safest therapy exercises for patients. It can be used in robotics, where robot assistants can be instructed to perform tasks through text instructions. In this research, a system was created to generate 3D animation of workouts from their textual instructions. An algorithm was developed to generate animation sequences from test cases. The algorithm utilized an animation graph which was created from a training set of 100 workouts with the nodes as postures and the links as animation-clip names along with their action data. The algorithm is designed to find and generate the closest animations available. A testing set of 40 similar workouts was used to test the algorithm and obtain an output of sequence of animations which were later depicted by the Unity Game Engine. In user evaluations, for 25 of the 40 test workouts, or 62.5% of the test workouts, the animation was determined by the users, to have the same or almost the same human motions as compared to a video of a human performing the workout. Analysis of the 15 workouts that were not similar to the human video (37.5%) showed that their issues can be fixed and the animation search can be improved by training the animation graph on more varieties of the workouts and on different text instructions of the same workouts from different sources. For 64.5% of the 156 sentences of all the text workouts, the animations were determined to correctly depict the text instructions. This research provides an initial animation graph and animation library which can be expanded to include more poses and animations from other domains such as yoga or dance. In the future system has the potential to provide users with no 3D modelling/animation expertise, a way to create 3D animations with just texts.
Date
2015
Document Availability at the Time of Submission
Secure the entire work for patent and/or proprietary purposes for a period of one year. Student has submitted appropriate documentation which states: During this period the copyright owner also agrees not to exercise her/his ownership rights, including public use in works, without prior authorization from LSU. At the end of the one year period, either we or LSU may request an automatic extension for one additional year. At the end of the one year secure period (or its extension, if such is requested), the work will be released for access worldwide.
Recommended Citation
Chimmalgi, Ravikumar Vishwanath, "Automated Conversion of Text Instructions to Human Motion Animation" (2015). LSU Master's Theses. 2198.
https://repository.lsu.edu/gradschool_theses/2198
Committee Chair
Knapp, Gerald
DOI
10.31390/gradschool_theses.2198