Machine Learning for Intangible Cultural Heritage: A Review of Techniques on Dance Analysis

Abstract

Performing arts and in particular dance is one of the most important domains of Intangible Cultural Heritage. However, preserving, documenting, analyzing and visually understanding choreographic patterns is a challenging task due to technical difficulties it involves. A choreography is a time-varying 3D process (4D) including dynamic co-interactions among different actors (dancers), emotional and style attributes, as well as supplementary ICH elements such as the music tempo, the rhythm, traditional costumes etc. Recent technological advancements have unleashed tremendous possibilities in capturing, documenting and storing Intangible Cultural Heritage content, which can now be generated at a greater volume and quality than ever before. The massive amounts of RGB-D and 3D skeleton data produced by video and motion capture devices. The huge number of different types of existing dances and variations dictate the need for organizing, archiving and analyzing dance-related cultural content in a tractable fashion and with lower computational and storage resource requirements. Motion capturing devices are programmable to extract humans’ skeleton data in terms of 3D points each corresponding to a human joint. This information can be combined with computer graphics software toolkits for modelling, classification and summarization purposes. In this chapter, we present recent trends in choreographic representation in terms of modelling, summarization and choreographic pose recognition. We survey recent approaches employed for the extraction of representative primitives of choreographic sequences, the recognition of choreographic pose and dance movements, as well as for the analysis and semantic representation of choreographic patterns.