Menu
The body language of dancers is an important tool for conveying emotions in the art of dance. In this study, the body movements of dancers are detected and tracked by Kinect, and the dance movement recognition model based on skeleton information and the dance movement emotion recognition model based on Attention-ConvLSTM are constructed respectively. After extracting and recognizing the dance actions, the emotion conveyed by each action is then recognized and measured. The action recognition accuracy of the dance action recognition model in this paper is 88.34%, which exceeds that of other action recognition models, and reaches the best result after only 40 iterations, and the recognition efficiency is also higher than that of other models. The emotion recognition model of the dance action emotion recognition model in this paper has the best emotion recognition effect, and the recognition rate is 98.95%. In the eigenvalue speed, the similarity between Excited and Pleased is the lowest (0.531). In Skeleton Pair Distance, the highest similarity between Excited and Pleased (8.439). For Skeleton Pair Inclination, the shortest distance between Excited and Relaxed is 0.503. For Speed, Emotional Relaxed, Sad and Miserable are easily confused with each other. For Skeleton Pair Distance, the emotions Pleased, Relaxed and Happy are easily confused with each other. For Skeletal Pair Angle, the emotions Mixed can easily be confused with the Happy category of emotions. Skeletal points of the legs have a greater influence on emotion expression. Finally, the construction of dance emotion expression mechanism is realized with the movement changes of head, hand, leg, waist and torso.