Contrastive Analysis And Visualization Of Dance Movement Based On Action Recognition

Dance is one of the oldest art forms of mankind and an important branch of ancient civilization.Dance originates from labor and is interpreted through human body movements,which express the dancer’s thoughts and eactions.With the gradual maturity of artificial intelligence and information technology,online education has become more popular,and more and more dancers have switched from traditional offline learning to online learning.However,for the special course of dance,ordinary online teaching is difficult to meet the guidance needs of dance teachers and the learning needs of dance practitioners,and online dance learning faces problems such as the inability to intuitively capture the accurate data information of dance movements through video alone,the difficulty of insight into the key defects of corresponding dance techniques and movements,the lack of corresponding interactive feedback,and the intuitive evaluation of dance movement quality.In response to the above problems,this paper proposes a method of comparative analysis and visualization of dance movements based on action recognition.Firstly,the three-dimensional coordinate information of key points of dancers in dance video is extracted through human posture recognition technology,and the multi-dimensional spatiotemporal data such as joint angle,joint point trajectory and so on of dance movements are obtained;Then,a human action recognition method based on hypergraph convolution is used to identify the dance action,which is compared with the standard reference action.Under the guidance of dance experts,the dance action quality is evaluated by the improved dance action evaluation model;Finally,a visual interface of dance action information based on multi-view collaborative layout is constructed to intuitively display the quantitative information of dance action and provide support for users to explore dance action information.Experiments demonstrate the effectiveness and feasibility of the method proposed in this paper.The main work and innovation of this dissertation include the following:1.In the action recognition method based on skeleton information,this paper proposes a new hypergraph convolution method to represent the relationship between multiple joints and dynamically refine the higher-order relationship between hyperjoints.An auxiliary module is proposed for the first time to realize the spatiotemporal modeling of the relationship between multiple joints and extract the effective spatiotemporal relationship of multiple joints.By comparing with other widely used model methods,the effectiveness of the proposed network model method is verified.2.In the comparative analysis of movements,this paper obtains the key frames of dance movements by using the method based on skeleton information and clustering,and realizes the similarity measurement of joint angles between the key frames of dance movements.Experiments verify the accuracy of key frame prediction;Then,based on the improved analytic hierarchy process,a comparative evaluation model of dance movements is constructed to obtain dance movement scores.The effectiveness of the evaluation model is demonstrated by experiments.3.In the dance action information visualization,this paper studies the advantages and disadvantages of the visualization method and the applicable scenarios,builds a dance action information visualization system based on multi-view collaborative layout,and evaluates the visual design through user research.