Optimizing user expectations for video semantic filtering and abstraction

Abstract

In this paper, we describe a novel automatic system that generates personalized videos based on semantic filtering or summarization techniques. This system uses a new set of more than one hundred visual semantic detectors that automatically detect video concepts in faster than realtime. Based on the personal profiles, the system either generates video summaries from video databases or filtered video contents from live broadcasting videos. The prototype experiments have shown the effectiveness and stabilities of the system. © 2005 IEEE.