Remote Dance Action Correction Based on EM and Min-min Algorithms

In the process of community learning for remote dance teaching, to ensure that students online master the correct dance actions and skills, and to improve the quality of teaching and students' learning outcomes, this paper proposes an action correction method in remote dance teaching based on the EM and the Min-min algorithms. This method takes the remote dance video as the foundation, first using principal component analysis (PCA) to segment the action sequence of the remote dance action, then inputting the segmented remote dance action into the HENet network model, and using the EM algorithm to estimate the parameters of the network. With the optimal parameters, the HENet network model is iteratively trained to extract the key joint features within the dance action. Then, a hierarchical posture extraction for dance action based on the relationship of the joint set is applied to extract the dance posture, Using the student's posture aligning the standard action, the dance action guidance vector is calculated, and the Min-min algorithm is used to allocate and schedule multiple guidance vectors, thereby achieving dance action corrections. Experimental results show that the segmentation accuracy remains above 88%, and the action accuracy is improved by about 20%, reaches accuracy between 88 and 98% and real-time (8–12 ms/frame). The extraction error of key joints is effectively reduced to between 1 and 2%. All these experiments verify the significant effect of the proposed method in improving the quality and learning effectiveness of remote dance.