Recognition method of dance rotation based on multi-feature fusion. (21st January 2022)

• Record Type: Journal Article
• Authors: Liu, Yang; Fan, Meiyan; Xu, Wenfeng
• Is Part Of: International journal of arts and technology. Volume 13:Number 2(2021)
• Page Start: 91
• Page End: 107
• Publication Date: 2022-01-21
• Subjects: multi-feature fusion -- mesh division -- directional gradient -- optical flow field; Technology and the arts -- Periodicals; Computer art -- Periodicals; 700.285
• Languages: English
• ISSNs: 1754-8853

Abstract

There are some problems in traditional dance rotation recognition methods, such as low accuracy of contour superposition and low recognition rate. A dance rotation recognition method based on multi-feature fusion is proposed. The background noise subtraction method is used to separate the human motion regions in the foreground of the video data, and the contour features of each frame image of the preprocessed dance video are superimposed to obtain the direction gradient histogram features of the dance action information. According to the law of optical flow, the feature vectors of the histogram of optical flow direction are normalised. According to the shape and motion characteristics of human dance in dance video, the dance rotation recognition classifier is constructed to complete the dance rotation recognition based on multi-feature fusion. The experimental results show that the proposed method has higher accuracy of 97% and lower error rate of 0.7%.