Multi-view RGB-D System for Person Specific Activity Recognition in the context of holographic communication

Abstract

Activity recognition is a key component of context-aware holographic communication to support optimal quality flow of data, but conventional approaches often lack in semantic information and context-awareness due to problems such as difficulty identifying the activity which the invidiual performs; overfitting when building activity models; collection of a large amount of labeled data from each end user. This paper presents a fully developed multi-view RGB-D system based on user-specific metrics - facial features coupled with a body skeleton. The system employs a skeleton-based approach. We test the performance of the proposed architecture in a controlled environment.