An intelligent subtitle detection model for locating television commercials

Area of Science:

• Computer Science
• Artificial Intelligence
• Signal Processing

Background:

• Television (TV) commercial detection is crucial for content analysis and advertising management.
• Manual identification of TV commercials is time-consuming and labor-intensive.
• Automated methods are needed to efficiently segment TV programs from commercial breaks.

Purpose of the Study:

• To develop an automated strategy for accurately locating television commercials within TV programs.
• To leverage subtitle absence as a primary indicator for commercial detection.
• To improve the efficiency and accuracy of TV commercial boundary identification.

Main Methods:

• A two-stage approach was employed: subtitle detection and commercial boundary localization.
• The first stage utilized six subtitle constraints and an adaptive neurofuzzy inference system (ANFIS) to identify frames without subtitles.
• The second stage employed a genetic algorithm to determine the mark-in/mark-out points of commercials, with an interactive interface for manual refinement.

Main Results:

• The proposed strategy successfully differentiates between TV programs and commercials based on subtitle presence.
• Experimental results demonstrated high performance with precision and recall rates exceeding 90%.
• The interactive user interface facilitated efficient boundary identification and correction.

Conclusions:

• The developed method provides an effective and accurate solution for automated TV commercial detection.
• The combination of subtitle analysis and genetic algorithms offers a robust approach to segmenting broadcast content.
• This strategy can significantly aid in media analysis, content management, and advertising research.