The use of artificial intelligence-based Siamese neural network in personalized guidance for sports dance teaching

This work aims to explore an accurate and effective method for recognizing dance movement features, providing precise personalized guidance for sports dance teaching. First, a human skeletal graph is constructed. A graph convolutional network (GCN) is employed to extract features from the nodes (joints) and edges (bone connections) in the graph structure, capturing both spatial relationships and temporal dynamics between joints. The GCN generates effective motion representations by aggregating the features of each node and its neighboring nodes. A dance movement recognition model combining GCN and a Siamese neural network (SNN) is proposed. The GCN module is responsible for extracting spatial features from the skeletal graph, while the SNN module evaluates the similarity between different skeletal sequences by comparing their features. The SNN employs a twin network structure, where two identical and parameter-sharing feature extraction networks process two input samples and calculate their distance or similarity in a high-dimensional feature space. The model is trained and validated on the COCO dataset. The results show that the proposed GCN-SNN model achieves an accuracy of 96.72% and an F1 score of 86.55%, significantly outperforming other comparison models. This work not only provides an efficient and intelligent personalized guidance method for sports dance teaching but also opens new avenues for the application of artificial intelligence in the education sector.