Humanoid Robot Behavior Emulation and Representation

• Main Authors: Yu-An Zheng, 鄭宇安
• Other Authors: Kao-Shing Hwang
• Format: Others
• Language: en_US
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/56516652039643393332

碩士 === 國立中山大學 === 電機工程學系研究所 === 101 === The objective of the thesis is utilizing body sensing technology to develop a more intuitive and convenient way to control robots. The idea is to build a body sensing control system based on Kinect framework. Through Kinect, users from different age groups can achieve the desired purposes through motion demonstration without complicated programming. The system can accurately calculate angle change from users’ gestures in a motion and identify key-postures which can compose an emulation motion similar to the presenting one. In other words, from analyzing these key postures, the demonstrated behaviors are able to be represented internally. Therefore, the system, consisting of a kinematics computational module and a representation algorithm, not only provides the function of behavior emulation but also behavior representation. By representation algorithm, the system extracts the features of combined behaviors. Besides, with the modular programming methodology, different behaviors can be reorganized to generate new behaviors based on the set of key poses represented by the extracted features. The application implemented in this system is within the OpenNI and NITE environment. OpenNI is used to retrieve information that the Kinect captured. NITE is used to track the user skeleton. The system is demonstrated by a play of “Tai-Ji-Advancer” and at http://www.youtube.com/watch?v=cSYS49JKVAA.