Dynamic Hand Gesture Recognition Based On 3D Skeleton

In this paper, we propose a dynamic hand gesture recognition method based on 3D Skeleton. According to the characteristics of dynamic gestures, we use the distance between different joints to represent hand shape feature, the movement of some certain joints to represent trajectory feature and the relationship between some certain joints to describe hand rotation feature. These three features are then input onto the SVM classifier. The proposed method is evaluated on a challenging dataset called DHG-14/28, experimental result shows that our proposed method is effective and outperforms start-of-the-art methods.