Extraction of Non Manual Features for Videobased Sign Language Recognition

Videobased sign language recognition is barely investigated in the field of image processing. General conditions like realtime-ability and userand environmentindependence require compromise solutions. This paper presents a system for automatic analyzing of the facial actions. For this point distribution models and active shape models are brought into action. Additional a comparison is made between different approaches for the shapes initialization. Thomas ~ z i u r z ~ k ~ Chair for Technical Computer Science Aachen University Of Technology Existing sign language recognition systems rely exclusively on the manual parameters [I] [2] [3] although image processing offers the possibility to consider non-manual parameters, too. This work presents a coarse overview of a system in development for the automated analysis of the human mimic. The extraction of eyeand lip features using a biomechanical model is described in detail.