Graphical models for multi-modal automatic video editing in meetings

Benedikt Hörnier, Dejan Arsić, Björn Schuller, Gerhard Rigoll

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

In this work we present a multi-modal video editing system for meetings, which uses graphical models for the segmentation and classification of the video modes. The task of video editing is about selecting the camera, that represents the meeting in the best way out of various available cameras. Therefore a new training structure for graphical models was developed. This is necessary for the learning of boundaries combined with the video mode itself. All developed and known decoding structures can be easily connected for an EM-training to our training structure. The achieved results of the system are state of the art.

Original languageEnglish
Title of host publicationDSP 2009:16th International Conference on Digital Signal Processing, Proceedings
DOIs
StatePublished - 2009
EventDSP 2009:16th International Conference on Digital Signal Processing - Santorini, Greece
Duration: 5 Jul 20097 Jul 2009

Publication series

NameDSP 2009: 16th International Conference on Digital Signal Processing, Proceedings

Conference

ConferenceDSP 2009:16th International Conference on Digital Signal Processing
Country/TerritoryGreece
CitySantorini
Period5/07/097/07/09

Keywords

  • Human-machine interaction
  • Machine learning
  • Meeting analysis
  • Multi cameras
  • Multi-modal low level features

Fingerprint

Dive into the research topics of 'Graphical models for multi-modal automatic video editing in meetings'. Together they form a unique fingerprint.

Cite this