Gesture Recognition and Generation for HOAP-2 Robots by Fuzzy Inference System

Since HOAP series robots resemble human body structure, a HOAP robot is expected to interact with others in real-time. However, it has proven hard in terms of learning, recognition, and interaction in real-time. In this paper a Fuzzy Inference System (FIS) is proposed, which learns gestures with segmentation and motion primitives, recognize gestures with created rule-based system in learning phase, and generate interactive gesture for HOAP-2 robot using real-world human interaction patterns. We also have a pre-processing element during gesture learning, which helps in better fuzzy rule generation and recognition with motion recognizer. Finally, at interactive gesture generation phase, some interactive parameters are incorporated to generate best possible response. We demonstrate the validity of proposed model with several interactive gestures of HOAP-2 robot in real-time.