Automatic human action analysis for human robot interaction (パターン認識・メディア理解)

概要

論文の詳細を見る
As recent interests in robotics move to intelligent robot and humanoid, human-robot interaction (HRI) technologies have become more essential and important. In case of the human being, they interact by recognizing various cues such as hand sign, face, gesture, speech, and so on. HRI can be accomplished by mimicking this way. Automatic recognition of human action which uses whole body action is required for HRI to communicate naturally and comfortably. This presents challenging problems, because detecting and tracking 3D human body components, and describing and modeling human action patterns from the body motion is a complex task. A human subject is firstly described by a set of features encoding the angular relations between a dozen body parts. Then, the extracted features are analyzed by HMM. For verifying the proposed method, we make several experiments using Korea University Gesture Database and apply technologies to interesting applications for HRI. The results and demonstration show that the proposed method can be effective in HRI, for automatic recognition of human action from motion sequences.
社団法人電子情報通信学会の論文
2007-08-27