Automatic description of human actions in the real world

2016 Fiscal Year Final Research Report

• Project/Area Number: 26540133
• Research Category: Grant-in-Aid for Challenging Exploratory Research
• Allocation Type: Multi-year Fund
• Research Field: Intelligent robotics
• Research Institution: The University of Tokyo
• Principal Investigator: Takano Wataru 東京大学, 大学院情報理工学系研究科, 准教授 (30512090)
• Project Period (FY): 2014-04-01 – 2017-03-31
• Keywords: 行動認識 / 機械学習 / 統計モデリング

Outline of Final Research Achievements

Artificial intelligence is expected to symbolically understand the real world. In this research, we have focused the human actions, and developed the fundamental framework to convert the actions to their relevant descriptions. This framework encodes human whole body motions and objects to be manipulated, and subsequently establish the connection between these symbols and the descriptions. This connection allows for generating sentences describing the human actions. Especially, the extraction of the contexts from sequences of the actions and sentences leads to selecting words appropriate to the current situations, and results in making correct descriptions for the human actions.

Free Research Field

ロボティクス