Research on Data Extraction of Tibetan Dance Movements Based on Different Convolution Models

Abstract

Abstract In order to strengthen the in-depth integration between culture and technology, and provide a certain theoretical basis for the improvement of the digital system of Tibetan dance and the innovation of Tibetan dance teaching, a new research method was proposed : based on high-definition Tibetan dance video data, we used OpenPose to extract the dancer's skeleton joint point data, and the five characteristics of Tibetan dance: "open", "tremor", "left", "round", "shun" and other characteristic movements were summarized. In the feature extraction stage, we tried different convolution methods to compare the extraction results. In the end, we extracted the dancer's bone point data more completely, deeply digitized the movement characteristics of Tibetan dance through digital technology. The experimental results show that the numerical feature rules of the five major characteristics are highly consistent with the dancer's performance analysis characteristics, which proves that the extraction and analysis method given is feasible.