Technologies and Applications for Next-generation Media

Sub-project 4 studies content-based multimedia retrieval and adaptive media playing technologies. In the past two years, we have developed content-based retrieval subsystems for image, video and music respectively. Besides, novel approaches for facial and body modeling are proposed and a text-to-animation conversion prototype system was implemented. Furthermore, a periodical multicasting protocol is proposed. Based on this protocol, a prototype video streaming subsystem is build. For the coming year, we will extend existing query functions to multi-modal query interfaces for respective media retrieval subsystems. Besides, we will also work on the integration of each subsystem. For retrieval from synthesis, we will study the classification technique to classify low-level animation coding into high-level semantic terms based on the frameworks of MPEG-7 standard.