An Adaptive Visual Gesture Based Interface for Human Machine Interaction in Intelligent Workspaces

This paper describes the implementation of an artificial neural network based adaptive hand positioning and gesture recognition system for multimodal interfaces in intelligent workspaces. The system is based on a stereo vision system and provides a series of mechanisms for an efficient segmentation of the arm-hand structure over the screen and its description in terms of moments that can be used by a neural network based structure in order to provide position and gesture classification results. These can be adapted to the user's desires through an interaction and feedback mechanism permitting a natural mutual learning process between the system and the human