Abstract: Simple and natural interaction between human robots is an attractive but challenging field in robotics.  Speech and gestures are two predominant ways for Human-Robot-Interaction.

This research work aims to develop a robust gesture recognition system for robotics applications.  The system will be able to work in cluttered and dynamic environment.  To achieve naturalness, non-intrusive cameras are used and no special requirements about clothes or markers are needed.  A model-based approach is employed and every joint angle estimated.  In order to have a comprehensive understanding of human motion and intention, both 3D spatial information of static gestures and meanings of dynamic actions will be recognised.  Moreover, face recognition techniques will be used to identify users, so that the system can deal with many-to-many mappings between gestures and meanings by different people.

This report emphasises the importance of gesture recognition, includes a brief literature review, presents the current research progress and proposes a detailed plan for future work.

About the Speaker: Zhi Li is a postgraduate masters research student with the Department of Electrical and Computer Systems Engineering