Research on original environment folk dance movement evaluation based on spatio-temporal graph convolutional networks

The original environment folk dance of China s ethnic minorities is an important part of the national intangible cultural heritage. However, cultural diversification has impacted the development and inheritance of original environment folk dance. Traditional dance inheritance is a one-way and artificial way, which makes it impossible for dance learners to get effective and timely feedback. Moreover, it is also crucial for dance trainers of different levels to choose suitable dance movements for practice. The human skeleton provides important information in the study of human movement. In order to better protect and inherit the original environment folk dance, we used a Kinect depth camera to record the skeletal information and to determine and repair occluded skeletal points in this study, and then adopted dynamic weight to improve the participation of key parts of the data. Combined with the spatio-temporal graph convolutional network, Evaluate the trainer s movements. We then verified this in the public dataset and in the original environment folk dance dataset that we recorded. The experimental results show that the numerical score of our method is positively correlated with the dance level of the subjects tested. It proves the potential of our method in the original environment folk dance movement-assisted practice.