"Core Strength" of Dance Lala Training Considering the Body Motion Tracking Video and Predictive Model
Jing Zhao · 2022 International Conference on Electronics and Renewable Systems (ICEARS) · 2022
"Core Strength" of the dance Lala training considering the body motion tracking video and predictive model is studied in this paper. When constructing the training sample template, it is necessary to set the value of the number of poses in advance. Too small a value will increase the error of the training action behavior template and reduce the recognition accuracy; too large a value will also increase the complexity of model training, training and recognition execution time will also increase. The function of the pose estimation module is to complete the description and expression of the specific posture of the human body at the current moment according to the frame image. After processing by the pose estimation module, each frame in the sequence can be calculated to obtain a feature vector, so that the image sequence can be converted into a feature vector sequence for the next step of action recognition. With these models, the applications on the dance Lala training is tested.