Profiling Balinese Dances with Silhouette Sequence Pattern Analysis

The Balinese dances are a very valuable asset for the tourism industry in Bali. The first objective of this study is to produce the first and standardized dataset of Balinese dance videos for the research of sequence pattern analysis of the digital image frames contained therein. The second objective is to implement and to evaluate the performance of our proposed scheme of feature extraction methods and Bag of Visual Movement on the pattern analysis. The third objective is to detect and to extract the patterns of Balinese dance silhouettes with quantitative analysis data in the field of computer vision. Results show that the Bag of Visual Movement generates visually differentiable clusters for image frames and our proposed method with HoG feature extraction and Bag of Visual Movement are visually able to detect patterns. The 13 pattern measures proposed are able to build profile of different Balinese dances from different maestros.