Extraction of dance features by unsupervised deep learning of motion capture data and application to education

2021 Fiscal Year Final Research Report

• Project/Area Number: 18K02893
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 09070:Educational technology-related
• Research Institution: Ochanomizu University
• Principal Investigator: NAKAMURA MINAKO お茶の水女子大学, 基幹研究院, 准教授 (20345408)
• Co-Investigator(Kenkyū-buntansha): 芝野 耕司 東京外国語大学, その他部局等, 名誉教授 (50216024)
• Project Period (FY): 2018-04-01 – 2022-03-31
• Keywords: 民族舞踊学 / モーションキャプチャ / Labanotation / OpenPose

Outline of Final Research Achievements

We collected motion capture data in the multimodal database of Carnegie Mellon University, which is available on the Internet, and investigated papers presented at international dance-related conferences to obtain motion capture data, but we were unable to collect sufficient data. This is due to the fact that motion capture requires a lot of time and effort.
In 2017, Open Pose, which obtains motion data by applying deep learning to video data from CMU, was developed, making it possible to obtain motion data from a large amount of video data available on the Internet.
The research method was changed to research utilizing these results, and a new proposal for Grant-in-Aid for Scientific Research(B） (21H03771) was made.

Free Research Field

民族舞踊学

Academic Significance and Societal Importance of the Research Achievements

現在先端研究分野で最も注目を集めているDeep Learningを用いて，これまでDeep LearningやAIが適用されることの殆どなかった動作認識分野を開拓する点がこの研究の独創的な点である。Deep Learningを同時に比較民族舞踊研究の視点から各民族舞踊の特徴抽出を行うことによって，舞踊のモーションキャプチャデータからの内在的分析を可能とすることによって，量的に基盤を持つ舞踊研究を可能とするとともに，より客観的な舞踊研究への道を拓く。