◎羅　　丹・Jun Ohya（早大）

The use of human gesture as a natural interface plays an utmost important role for achieving intelligent Human Computer Interaction (HCI). This paper presents a visual-based gesture recognition framework, which combines different groups of features such as facial expression, hand motion and hand shape. We employ two fusion strategies in the feature level and decision level to the combined multi-modal features. An adaptive CRF and condensation-based algorithm is adopted for classification. Experimental results showed that facial analysis and hand shape information improved hand gesture recognition and decision level fusion performs better than feature level fusion.