Closed Captioning Quality Assurance

AT&T is a fully integrated solution provider that delivers advanced mobile services, next-generation TV entertainment, high-speed Internet and smart solutions for people and businesses. Our mission: Connect people with their world, everywhere they live, work and play … and do it better than anyone else.

AT&T video services deliver an audio visual experience to every customer. Closed Captions provided along with our video content enables people with difficulty hearing the audio to still watch and enjoy our programs. Occasionally, there can be errors in the closed captions such as missing text, garbled text, delay between the audio and the captions, etc. The errors may be in the source video that is provided to AT&T or it may be in the encoded video that is delivered from AT&T to our customers. In this project we are interested in detecting and analyzing errors that may occur in closed captions to ensure that what is provided to the customers is of the best possible quality.

The project will involve developing a tool that AT&T will use to analyze closed captions in a video file. The tool will be able to detect and decode closed captions according to the captioning standards that are used by AT&T. The decoded captions will be displayed alongside the video and audio in order for an operator to view the rendered captions and determine if they are correct. In addition, automated tools will be provided to detect errors such as garbled closed captions, missing text, and timing errors between captions and audio.

 

Project Statement

Students will develop a closed captions analysis tool that will have the following features:

·         Decode captions and video/audio from a video file in order to perform further analysis.

·         Render the captions to an image or video sequence with appropriate timing, font, text color, background color, etc.

·         Automatically determine if captions have missing text.

·         Automatically determine if captions are garbled.

·         Automatically determine if closed captions language is different from language specified in the metadata.

·         Simple to operate with an intuitive GUI.

 

In addition to the above, a stretch goal for the project would include the following feature:

·         Render closed captions in real-time as an overlay on the source video and audio with accurate timing, font, text color, background color, etc.

 

Deliverables:

        Final deliverables at the end of the clinic shall consist of a detailed report documenting all findings (problem areas, solutions, ideas, etc), and a software deliverable that includes the key features detailed in the Project Statement. Additionally, two presentations (mid-year, and final) shall be given to AT&T staff which outlines the status of the project, key problem areas, and next steps.  At the end of the project, the students should have an understanding about closed captioning systems in video, speech detection, recognition and classification, and user experience design.

Meeting time: Fridays 11am-12pm.

Quick Links:

Student Team
  • Christian Cano
  • Benjamin Jennings
  • Gevorg Khachatrian
  • Angie Sanchez
  • David Yardimian
Project Sponsor
Project Liaisons
Faculty Advisors
Resources