TRL
TOP PAGETokyo Research LaboratoryEmploymentProjectsRelated InformationIBM Research
Japanese page is here.

Video digest generation system



Overview

Starting with original video sources, this system aims at creating and distributing individualized video digests based on user profiles. Each digest is generated based on two kinds of profiles, user profiles including user's preferences, and a recommendation profile defined by the content provider. The importance of each event in the original video is calculated using the metadata about the video source and the two profiles, and the video digests are generated dynamically. Each user can also specify the digest length.

Video digester

Video digester calculates the importance of each event from the MPEG-7 metadata and the profile information, and generates a digest description in SMIL. The generated digest information is transmitted to a video delivery server. Each video digest is generated dynamically based on the SMIL description, and then distributed.

VideoDigester

Profile

To guide the digest generation process, profiles are generated with the importance of each event described using MPEG-7. There are two kinds of profiles, one describing recommendations from the content provider, and the other describing the users' interests and preferences. The recommendation profile is actually defined for typical groups of viewers. For example, in a sport the fans of one team would be a group with a specified recommendation profile. The system can simultaneously combine both of the profiles in the process of digest generation.

@
Profile


Research home IBM home Order Privacy Legal Contact IBM
Last modified 26 Feb. 2002