Research Article
A Method of Recognizing Specific Movements in Children’s Dance Teaching Video Based on Edge Features
@INPROCEEDINGS{10.1007/978-3-031-50552-2_14, author={Chunhui Liu and Chao Long}, title={A Method of Recognizing Specific Movements in Children’s Dance Teaching Video Based on Edge Features}, proceedings={Advanced Hybrid Information Processing. 7th EAI International Conference, ADHIP 2023, Harbin, China, September 22-24, 2023, Proceedings, Part IV}, proceedings_a={ADHIP PART 4}, year={2024}, month={3}, keywords={Edge Feature Children’s Dance Teaching Video Pretreatment Adaboost-BP Neural Network Specific Action Recognition Method}, doi={10.1007/978-3-031-50552-2_14} }- Chunhui Liu
Chao Long
Year: 2024
A Method of Recognizing Specific Movements in Children’s Dance Teaching Video Based on Edge Features
ADHIP PART 4
Springer
DOI: 10.1007/978-3-031-50552-2_14
Abstract
Dance videos have issues with self occlusion and high complexity of actions, which affect the effectiveness of action recognition. In order to improve the accuracy of action recognition, a video action recognition method for children’s dance teaching based on edge features is proposed. Image preprocessing for children’s dance teaching videos, including grayscale and enhancement of video images. The background subtraction method is used to detect moving objects in video images, and the Canny operator is used to detect the edges of moving objects, enhancing the continuity of the edges. After obtaining an image that only includes the edges of the object, further extract the contour features of the object. A recognition method based on Adaboost BP neural network has been constructed. Using the BP neural network as a weak classifier, the Adaboost algorithm is combined with the outputs of multiple BP neural networks to construct a strong classifier, avoiding falling into local optima. Using edge features as input to achieve specific action recognition for children’s dance teaching videos. The experimental results show that the recognition method based on edge features has a high average recognition accuracy of 94.715%.
