The aim of this project is to undertake research into computer vision-based, human-computer interaction for the disabled using gesture recognition. This research aims to further the state of the art in gesture recognition by developing techniques for on-line training and adaptation of gesture models and their use for robust, real-time, dynamic recognition of two-handed gestures. Interactive training will allow users, particularly those with motor impairment, to develop and use their own gesture vocabulary. The aims of the project are:

(1) To investigate and develop computational methods for real-time, dynamic recognition of two-handed gestures based on robust tracking of the head and hands.
(2) To develop an interactive system for on-line training of gesture models, along with the necessary computational methods for incremental model building and adaptation.
(3) To investigate the application of these techniques to gesture-driven Human-Computer Interaction for disabled people.