A two-level approach for modeling and recognition of hand gesture

The main characteristics of human hand gestures can be summarized by their dynamic, multi-attribute properties. Previous research on gesture pattern interpretation has shown the possibility of recognising local aspects of hand gestures. Nevertheless, a global framework for interpreting the whole gesture from local aspects has yet to be provided. We propose a two-level approach using colored Petri nets interlinked with hidden Markov model based low-level continuous recognition. This approach provides an integrated modeling and recognition capability for hand gestural interaction.