A confidence based recognition system for TV commercial extraction

  • Authors:

  • Yijun Li;Danqing Zhang;Xiangmin Zhou;Jesse S. Jin

  • Affiliations:

  • Nielsen Media Research, Australia;Queensland University of Technology, Australia;University of Queensland, St Lucia Campus, Australia;University of Newcastle, Australia

  • Venue:
  • ADC '08 Proceedings of the nineteenth conference on Australasian database - Volume 75
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

Automatic real-time recognition of TV commercials is an essential step for TV broadcast monitoring. It comprises two basic tasks: rapid detection of known commercials that are stored in a database, and accurate recognition of unknown ones that appear for the first time in TV streaming. The existing approaches, however, can not perform robust commercial detection because they highly rely on the assumption that black frames are inserted before commercial breaks. In this paper, we propose a novel confidence-based recognition system to address this challenging issues. Those known commercials are detected by applying a fast search algorithm to a pre-built commercial database, and those unknown ones are determined by managing a buffer containing possible yet unconfirmed commercial spots. A new concept of confidence level is defined for each candidate, therefore commercial breaks can be accurately determined based on a confidence threshold. Experimental results using two 48-hour TV broadcasting indicate the high performance of our proposed method.