An Approach for Conceptual Modeling and Extracting Expressive Semantics from Dance Videos

摘要

Due to the high complexity and exponential growth of videos, video data management tasks, such as understanding and annotating semantic content, are very difficult to be accomplished. Dance videos are one such critical domain. Considering the various dance components, understanding their semantic content conceptually, extracting the expressive semantics, annotating them properly and storing the annotations in a suitable medium for future reuse and analysis, are the essential tasks for the next generation dance systems. This paper presents a conceptual multimedia data model based on Protocol Analysis, called Video Semantics Directed Acyclic Graph (VSDAG) model. The expressive semantic content, represented by VSDAG model, is extracted from the video by a unique video annotation system. The generated annotations are stored in a XML compliant structure.