抄録

H-045
表情と手振りの組み合わせから成る手話動作認識のためのマルチモーダル特徴量抽出法の検討

The use of gesture as a natural interface plays an utmost important role for achieving intelligent Human Computer Interaction (HCI). In this paper, we describe a two-stage method for extracting multimodal features, including facial expression, hand motion and hand shape features which are extracted from image frames. The first stage uses Modified Census Transform (MCT) based detector to propose face and hand position using Adaptive GMM. A second stage we combine the DCT based facial feature and hog hand shape feature with hand location temporally, which are dimensionally reduced.