Extraction of key postures from 3D human motion data for choreography summarization

摘要: Modelling and digitizing performing arts through motion capturing interfaces is an important aspect for the analysis, processing documentation of intangible cultural heritage assets. However, existing modelling approaches may involve huge amounts information which are difficult to process, store analyze. To address these limitations, usually a skeleton describing dancer extracted. often complexity still remains due high spatio-temporal dependencies detected joints. In this paper, alternative approach presented: choreography summarization. This means that very small number image frames extracted represent choreography, thus significantly reducing storage complexity. our problem summarization treated as unsupervised clustering approach. Evaluation indices introduced monitoring performance. Experimental results on real-life dancing performances verifies capability proposed method capture main patterns with trajectory points.