3D Digital Model of Folk Dance Based on Few-Shot Learning and Gesture Recognition

Abstract

Folk dance is a very unique local culture in China, and dances in different regions have different characteristics. With the development of 3D digital technology and human gesture recognition technology, how to apply it in folk dance is a question worth thinking about. So, this paper recognizes and collects dance movements through human body detection and tracking technology in human gesture recognition technology. Then, this paper writes the data into the AAM model for 3D digital modeling and retains the information by integrating the manifold ordering. Finally, this paper designs a folk dance learning method combined with the Few-Shot learning method. This paper also designs a data set test experiment, an algorithm data set comparison experiment, and a target matching algorithm comparison experiment to optimize the learning method designed in this paper. The final results show that the Few-Shot learning method based on gesture recognition 3D digital modeling of folk dances designed in this paper reduces the learning time by 17% compared with the traditional folk dance learning methods. And the Few-Shot learning method designed in this paper improves the dance action score by 14% compared with the traditional learning method.