@article{1618,
  author = {Arsalaan Ahmed Shaikh, Hammad Qureshi},
  title = {An Efficient Algorithm for Automatic Television Broadcast Monitoring},
  journal = {Signals and Telecommunication Journal},
  year = {2014},
  volume = {3},
  number = {2},
  doi = {},
  url = {http://www.dline.info/stj/fulltext/v3n2/2.pdf},
  abstract = {The last two decades have seen an unprecedented rise in the number of terrestrial cable and satellite TV channels. These channels generate a huge amount of digital content and are the main medium for businesses to advertise their products in a globalized world. Broadcast monitoring is an essential activity which involves evaluating whether the correct content was aired and for the correct length of time and at the time previously agreed upon. As advertisement forms the bulk of the revenue of the television channels, monitoring of advertisements becomes extremely important. Advertisers demand efficacy in the airing of their content and payments are made only when the claims are verified. Lack of efficient and inexpensive monitoring technologies directly impacts growth in advertisement revenues. Monitoring is a tedious task
which involves monitoring and analysis of thousands of hours of audio/visual content forming terabytes of data. Hence,
efficient automated techniques are required for an industry which relies mostly on manual auditing. Heterogeneity in aired advertisements, signal quality, sizable amount of multimedia data, and demand for high accuracy makes it a challenging problem to solve. In this paper, we propose a set of audio features which may be used to perform automatic auditing of broadcast content. A feasibility analysis is conducted based on a proposed Average Dependency &
Minimum-Maximum Distance criterion that judges the ability of a feature to differentiate between classes of advertisements. The criteria is combined with sequential floating search method to obtain an optimal or near optimal feature subset. The efficiency and effectiveness of audio features is evaluated on a real-world dataset of 216 hours of television broadcasts and 28 classes of advertisements. The results show that Gabor Filter Bank Feature (GBFE), Mel-frequency Cepstral Coefficient (MFCC) and MPEG7 Audio Flatness Mean (having yielded overall recognition rates of 99.33, 98.99 and 97.31 respectively)
are the most suitable audio features for an automated television broadcast monitoring system.},
}